View Single Post
  #26 (permalink)  
Old 11-15-2007, 12:45 PM
APOLAUF's Avatar
APOLAUF
PPCGeeks Regular
Offline
Location: Nashville, TN
 
Join Date: Oct 2007
Posts: 63
Reputation: 35
APOLAUF is just getting started
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
What this thread lacks is some hard numbers. I've run some TCPMP benchmarks on a clip that I normally use to benchmark.

I'm running DCD's clean ROM v1.4.5, all radios OFF.

Without performance CAB, after reboot:

Clip throughput is 1.0Mbit/s.

WITH performance CAB @"624"MHz, after reboot:

Clip throughput is 1.0Mbit/s

Change: 0.0 Mbit/s.

It does somehow, I don't know why, seem a bit snappier. Perhaps something was changed OS-wise, but NOTHING has changed as far as CPU performance. I have it set to "624" MHz. Let's see why nothing has changed: If this CAB is designed for an XScale CPU, the PXA270, which runs at 416MHz, there is no way in hell that a completely different processor, compatible ONLY because it belongs to the same ARM11 classification, will exhibit changes. The PXA270 is a unicore processor with Intel's power management and dynamic speed scaling built in. The MSM7500 has a Qualcomm ARM11 implementation with an ARM9 second core for phone functions. Remember, ARM sells intellectual property, not processors. Intel (well, now Freescale since they sold their XScale processor division) and Qualcomm have their own take on the processors, their ultimate power management, and speed stepping. They adhere to the ARM instruction set architecture. Therefore, these are completely different chipsets, though they in part share an instruction set architecture, and for that same reason, the CAB will do nothing for MSM7500. The benchmarks prove it.

Now, can we PLEASE set this to rest?
Reply With Quote