Out of curiosity...How are you measuring "performance"?
How much speedup you will see depends entierly on what you are rendering. With 5.0.2 we have started optimizing for Fermi cards and are seeing up to a two times speedup over the previous generation of cards for many filters. Running an individual filter two times faster will not however translate to doubling the speed of a render since the performance of many other pieces has remained constant. Reading frames from disk, uploading frames to the GPU and many other aspects do not change with a newer GPU.
As Steve has said unless you update to 5.0.2 you will not see a performance difference. Also as Shooternz said, unless you measure your performance with something like PPBM5 you may not know what your performance really is.
Hello my Adobe friends, I love Adobe so don't get me wrong here:-)
This is my system:
Intel i7 920
Asus P6T Deluxe V2 (X58 chipset)
Corsair 12GB (1600MHz)
Adaptec 3085 raid card with 4 x 1000GB Seagate in a raid 5
Nvidia GTX470 with the latest driver
AJA Kona LHi
C disk Intel SSD G2 with Windows 7 Ultimate 64bit with latest driver
Adobe CS5 with al the latest update inkl 5.0.2.
Just read this from dvinfo Adobe and our friend Harm Millaard, so I can not be that wrong about it?
"Currently MPE is still in it's infancy and CUDA Toolkit 3.1 has just appeared.
What does that mean?
First of all that at this moment in time there is no performance difference noticeable between a GTX-285 and a GTX-480, tested on the same system. This despite more than double the cores and double the width of the memory bus of the 480 over the 285.
At this moment the conclusion may be that the 460 is the most attractively priced MPE card with the same performance as the 480.
However, it is to be expected that when MPE get's more mature, it will use available cores more efficiently. Just like the threading on certain processes in CS5 that is far from optimal, causing much more latency in the communication between CPU - RAM - GPU - VRAM and back, especially on hexa cores.
Threading is a major factor for performance gains, both in the CPU and in the GPU. Therefore I expect that in time the 470 or 480 will be a much better choice than the 460".
Keep in mind that the post you quoted was written on August 10, and if you now take Steve Hoeg's response here, you can derive that Adobe is constantly at work to further improve MPE performance gains, so we are all shooting at a moving target, because nVidia is also constantly improving their drivers and CUDA support.
Bill Gerhke is one to have run the PPBM5 benchmark test with both a 480 and a 285 and his results did not differ enough to say that it is anything more than measurement errors.
We still haven't seen benchmark results that suggest that in 5.02 the infancy stage has been left behind, but knowing that Adobe is constantly improving its MPE code, I think it safe to say that in CS Next we will see the performance of the Fermi range of cards increase further, it is just a matter of time.
BTW, after reading my post again, I still think it is a valid post, despite the nearly three months that have passed since then.
Harm Millaard wrote:
Bill Gehrke is one to have run the PPBM5 benchmark test with both a 480 and a 285 and his results did not differ enough to say that it is anything more than measurement errors.
I would like to amend one thing above that were my conclusions with 5.0.1. With 5.0.2 here are some results:
Board AVI MPEG2-DVD H.264
Software MPE Total GTX 260 67 36 68 5 71 176
63 33 66 5 71 167 GTX 480 66 27 64 5 71 162
As you can see both the MPEG2-DVD encoding time and the H.264 encoding time had slight improvements as the capability of the board increased.
Though certainly not what we might expect given the specs of the cards and their price difference.
Not at all, it just shows the GPU isn't the bottleneck in these particular tests.
I don't get that. Based on the data supplied, it could just as easily show that the superior card isn't being fully utilized.
Thats the point with this thread "superior card isn't being fully utilized", the question is why?
There are two sides to this coin.
1. The GPU does all it can and is waiting for other components to finish, in which case the limited multi-threading and especially for the HT cores, not so much the physical cores, becomes a point of interest to further improve performance.
Now if this is indeed the case I find it hard to explain why both the CPU load and the GPU load remain below 50% during direct exporting and less than 70% of memory is used. OTOH, using AME increases peak CPU load to around 80% and GPU load to around 50%.
Anyway in both cases, direct export and AME, there appears room for performance improvement, since none of the components are being utilized for 100%.
2. It may be the case, as is rumored on the web, that not all CUDA cores are used, but only a fraction, in which case a peak load of only around 50% makes sense, but would also be a prime target of interest for increasing performance.