1 person found this helpful
Yes. All other factors being equal (memory, clock speed, etc.), the number of cores is the determining factor for performance.
Here are the results of my testing with our Premiere Pro BenchMark (PPBM5) different graphics cards in my fast but non-overclocked system You can see that the number of cores does improve performance but can decide for yourself the cost/performance advantages The cost numbers are now not current, those numbers were as of the middle of April this year. Also you can see that overclocking the GPU does not improve the results.
Is the number of core a important factor for AVCHD processing ?
That depends on what you mean here. Neither card will help you play back AVCHD media better. That get's handled by the CPU alone. If your system is having trouble playing AVCHD, you need a better CPU, not a better graphics card.
OK Bill and thank you very much for your datas and your answer !!!
Thank you also to Jim & Harm.
I just would like to give some precisions about my concern and how to understand your table.
In your table :
- Colomns "MPEG2-DVD" (yellow one ) : is it the output format you've given as an example ?
- What "Render time" line colomn signification on your table ? : is it the time necessary to render your job ?
In fact I would like to be able to add multiple effects, lot of layers in real time on After Effets CS5 with AVCHD file.
To drag some AVCHD native MTS file on my time line add some effects and layers... and would like the playback to be very very smouth.
2a./ I would be able to jump into the middle of the time line and the play back to be smouth also
2b./ and of course I would like to have a render time, or export time as small as possible
Can you see some answers to my concern in your PPBM5 results ?
Is the number of CUDA cores an important factor for this need ?
4./ In YOUTUBE you will fine what I would like te able to do by choosing the good GPU : GTS450 or GTX460
Many many thanks for your help.
I guess it is rather evident from your questions that you are not familar with Harm and my Premiere Pro BenchMark (PPBM5) which I refered you to earlier. This is the downloadable benchmark program that was used for these tests. Lots of effects many layers and and extensive table of results
lot of layers in real time on After Effets CS5 with AVCHD file.
After Effects does not take advantage of CUDA processing. Again, neither card will help with that. You need a faster CPU, and LOTS of RAM.
Actually, the GTS 450 has 192 CUDA cores (the same number of cores as the original revision of the GTX 260), not 128 CUDA cores. The predecessor of the GTS 450, the pre-Fermi GTS 250 (which actually started life as the 9800 GTX+, and based on the exact same G92 GPU as the GTS 250), had 128 CUDA cores and slow GDDR3 memory while the GTS 450 uses faster GDDR5 memory.
As a complement, I did an extact of the PPBM5 concerning 2 motherboard :
P67 and Z68
(Attached file is the excell sheet with the extract datas)
Same processor : I7 2600k
Same operating system : Win7 64
Same RAM : 16Go
It seems to be difficult to identify what is the performance key driver in fact...
Z68 seems not to be a good choice but why ?
Personal Computer Id Mlacom test OC
Z68X-UD5 GTX 580
MPE gain : 2,7x
Personal Computer Id Novak 470 Asus P8P67 EVO GTX 470
MPE gain : 16,5x
All being equal, do you thing if I take the configuration of "Novak 470" (Personal Computer Id) and only change the motherboard, I will obtain better results ?
I was sound that Z68 was better than P67 for video rendering
I'm a little bit confused...
httpppbm5.comBenchmark5.xls 589.0 K
Oups... thanks for this infos I didn't had !!
If you can give your feeling about my motherboard infuence (see my previous message) Ii will be very pleased :-)
Thank you for all of you for your quick answers you've already provided !!!!
I'm new to video editing and CS5.5. but you want as much RAM as possible. Ram is cheap. Also RAID hard drives really do a lot. As far as Cuda, the more the beter, although if you look at the bench marks (research FAQ at the top of page) there are plenty of Cuda graphics cards that do fine. Also check out http://forums.adobe.com/thread/810750?tstart=0 as well as some of the other RAID threads. You need to do a lot of other reseach but google is free (sorry). Y9ou seem to have more expreince than me but you want to be more specific about what you have and what you want to achieve. I'm guessing this will be more RAID and RAM. You can get plenty of Cuda on the cheap for a graphics card. RAM is cheap and RAID can be used for next computer as well. Just a thought.
You should be careful making your comparisons. We cannot weed out all the bad results that we get.
1. When you see an MPE-On results of 27 you should disregard those results entirely.something is definetly wrong. With an appreciable overclock anything much over 10 indicates something is wrong with their setup.
2. The other major problem with the Miacom results is that they were done with PP 5.0.0. It was a terrible scoring version. Also do not compare 5.0.3 results with 5.5.0 results as Adobe changed the caching algorithm and MPEG2-DVD scores are up to three times higher in CS5.5
3. We cannot tell frrom your spreadsheet
3.1 How much ram is used..
3.2 What the clock speeds are.
3.3 What the disk subsystems are.
Yet, apparently you picked those two systems for comparison based on the total scores.
4. Z68 are much newer boards and there are not as many for comparison purposes and most importantly since they are so new they are running CS5.5 which as I said in #2 above distorts the scoring.
5 What I just did in our posted results (so you can duplicate it) is to first sort on i7-26xx and then sort on CS5.5 and you will see (out of today's 17) that the top three results are Z68 motherboards!
5. You definitely do not want to make a board decision based on MPE ratio numbers, they are only significant to demonstrate that there is a gain in MPE performance
6. I personally recently bought a z68 board for my second edit station (mostly for experimental usage).