My system (*) shows a problem in the Time Line rendering time(s), in software as well as in MPE-CUDA. ( however, the other three PPBM5 tests are very acceptable, see underneath ).
I tried all possible 1, 2 , 3 , 4, 5 disk configurations ( as per Harm's Disk Setup Guide ) but in none of them I got an acceptable result for the Time Line rendering time.
I have been experimenting with this problem for many days now and I am about to give up.
In " PPBM(5) benchmark.pr " the Disk I/O , the MPEG2.DVD and the H.264 time are quite acceptable, i.e. at best resp. 73, 30 and 78 sec
However, the Time Line render times are very disappointing, i.e. at best 120 sec SW and with CUDA 40 sec.
Most peculiar is that, after slowing down the CPU from 4,2 GHz to ~3 GHz, the software time line rendering time increases obviously considerably but the CUDA rendering time maintains its low 3:1 ratio, while with a lower software rendering time, rendering with CUDA should produce a better ratio than the said 3:1 .
Obviously, in my system, something is wrong with the GTX 40 CUDA performance, but also with the software time line rendering time.
As said above, I emphasize, in contrast to this, the other three PPBM5 test results are fair enough ( at least for my system ).
I must assume the PCI-e highway CPU-GPU and vice versa is OK.
MY PCI-e GPU card GTX 460 @ 960 MHz ( 1GB ) ranks very high within the 460 benchmark class of the PassMark en Furmark GPU tests ( I reproduced these tests several times ).
Also my CPU I7 950 @ 4,2 GHz ranks high ( highest but one in the PassMark test ) within the i7 950 benchmarks class.
The 24 GB memory speed also tested very well in the real time benchmarks.
So, I wonder, if the CPU and the GPU and the PCI-e as well as the 24Gb RAM all work OK , what else could be the reason for the slow Time Line CUDA render times ( having configured the various disks cf Harm's Guide ).
Obviously, my GTX 460's CUDA does do something, but no better than 3 times the ( slow ) software rendering time.
And even the software rendering time itself is also slow ( at best 125 sec in PPBM5 Benchmark.pr ).
If necessary I am ready to replace another GX 460 or even to build a new system with another mobo, but I am afraid, with respect to the abovesaid, this will not cure the problem and I would regret to waste unnecessary money and time.
So, what else could I investigate to improve the Time Line rendering performance.
Although Premiere ( and AE as well ) runs very nice compared to my old system, I definitely do need a better CUDA timeline rendering ( CUDA was the main reason for me to buy my present system ).
Comparing to the PPBM5 Benchmark results list, my system's Time Line MPE-CUDA rendering should be in the 7 sec. range ( such as several i7 950/GTX460 systems on the PPBM5 Benchmark list show ) instead of my 40 sec now.
Any help would be greatly appreciated.
- W7 ultimate
- CS5 Premiere 5.0.3
- i7 950 @ 4,2 GHz ( 21 x 200 ), 24/7 stable,
- GTX 460 @ 960 MHz ( MSI Hawk ), 24/7 stable,
- 24 GB RAM Kingston 1600
- 3 x Raptor 300 GB in RAID 0
- 2 x 1 TB disks
- mobo ASROCK X58 Extreme 6
- Cooler Master Silent Pro 1000 W