I see no reason why it will not work as it has the required minimum amount of memory. If I win the lottery I will go out and buy one, or if you want to send me one I will be glad to test it for you. And yes, our beta benchmarks for the Fermi class boards roughly show a 10x factor improvement (this is on high end CPU and memory systems) on rendering the timeline, your mileage will vary with your own system!
Do not forget that Adobe is rumored to be hard at work optimizing the CUDA technology and I would not be surprised to find in a later release that the more cores and more memory will pay off big time.
Agreed Bill. Currently the MPE is in it's infancy and many effects still have to written in CUDA. The more effects that are, the more the card's Cores and Ram Frame Buffer(ie amount of ram on the card) come into play. Also at some point, I am sure the codecs as well will be written in CUDA but that will likely be the last since each codec will have to be translated. Right now the the majority of the load and threads have to be dispatched by the CPU to the GPU when it comes into play and sent back to the system Ram after the GPU does it's work. This is slowing the process down and limiting the usage of the video card. As more is added to the GPU processing list, the GPU specifications will effect performance exponentially since latency from the multiple device communication reduces.
I just installed a 1gb GTX 460 and activated the MPE "hack" as per instructions. Tech reviews of this GPU are stellar and I was eager to see MPE in action.
I ran a few performance tests with multiple layers, color correction, other GPU accel effects and saw no difference with GPU accelleration activated or software MCE only.. I used preview smoothness, scrubbing, and CPU usage as performance measures. I did one test render and saw no difference in speed.
I have a I7 860 with 8gb RAM and EVGA mobo.
Am I missing something? Is there a best way to test to see how the GPU is being utilized.
What type of material did you test with. How many layers? What effects did you use?
Did you run the GPUsniffer.exe to make sure the card was showing as supported after the change?
1 person found this helpful
The $ 200 GTX-460 has about the same number of cores as the new $ 1,200 Quadro 4000.
Talk about bang-for-the-buck. To see the impact of MPE acceleration, try the PPBM5 Benchmark
1 person found this helpful
Thanks for the benchmark. It revealed substantial gains in rendering.
After getting familiar with testing procedures, I ran a project with 2 layers of 1080p, color correction, and 3d effects. Turning off GPU MCE, the red bar appeared above the timeline and previewing was ineffective. Activing GPU MCE switched the bar back to orange, previewing and scrubbing were operational again. Rendering time within PP was about 10x faster. Exporting time was about 8x faster.
Success with the GTX 460. 1gb, gigabyte.
Glad to hear of your success
According to Anandtech, Gainward is going to put out a 2GB GTX460 card in the near future for about $280.
In the gaming benchmarks, jumping to 1GB from something smaller often yields a nice return. But doubling that to 2GB hasn't shown the same type of improvement, at least in the benchies I've looked at from prior generation cards.
I'm curious what effect having 2GB for editing apps might mean. The GTX260 is already shaping up as one of the best all-round launches of the current era. Great for gaming, plenty of CUDA cores, highly scalable on SLI, lower power draw and not as hot or loud.
My guess is that Gainward has determined it can squeeze an extra $30 or so out of a niche market for the gamers who have seen that extra memory helps in massive resolution triple monitor gaming. Whether or not the extra GB will actually make much difference in gaming remains to be seen, but as a marketing ploy it seems bankable enough.
It would be really cool if we can ride this gimmic'ed card into the promised land. $500 cards aren't going to be the solution for all that many CS5 users. I haven't been reading 'round here for the last 3 weeks all that much, but do recall a post detailing that the GTX480 can't be easily driven to it's full capacity. Seems it's almost too powerful.
If a GTX 460 with nearly three-quarters of the stream processors of the GTX480 and even more memory can be had for about half as much, then this might shape up to be a neat li'l coup for the editing crowd.
Wonder if anyone will ever run a comparison of the three memory variants from the CS5 editing point of view?
You know how to lie with statistics? Well, I'm a firm believer in large number tests, like PPBM5 Benchmark
It may not immediately give you all the answers you are looking for, due to the limited number of submitted results, but as the numbers grow, you can derive very valuable information from the results. As the number of results increase, the statistical validity goes up and its value too.
Tests like Tom's Hardware, that were linked to here, have several distinct drawbacks, in that you can't compare your own system with their results and because they never explained what they tested, so repeatability is a major issue. PPBM5 Benchmark does not have those drawbacks. However, to make the test more valuable for everyone, results on individual setups must be submitted for inclusion.
As Bill and I get more results, we will update our conslusions and background info as well.
Currently we have no 460 results yet, so consider this an invitation to submit your results, especially if you have a 460.
I like the look of the New PPMB5 site.
Its getting better and better all the time.
Thanks for keeping it going.
That info really helps.
On a side note.
On the PPBM4 site if you click on the Mac Problems page and from there try to click BENCHMARK RESULTS.
The link goes to a wrong page since the hyperlink is wrong.
this is what it points to:
that extra ppbm4 throws it off.
Anyway thanks for all the work you and Bill are doing.
Thanks for your feedback. Appreciated. I do hope Bill reads it as well, since then he can correct the PPBM4 site, while I keep my attention focussed on the PPBM5 site.
Thanks for the benchmarking program. Here are my output files. Comparing my results to data on the ppbm site, my system seems pretty dissapointing, though I'm happy with improvements with the GTX 460. I've tried overclocking my i7 860 on an evga p55 micro, with no success.
Your data were not submitted in the regular way, so they are pretty hard to check, but the results make no sense at all.
MPEG of 38 would rank as a major miracle running on steroids. My guess is that it should be 238, comparable to other systems with almost identical hardware above it. Same with your H.264 score, 71 secs should probably be 171. Your AVI score is a disaster.
I ask you to open AME, create the MPEG2-DVD queue, press duplicate 3 times, create the H.264-BR queue, press duplicate 3 times, create the Disk queue, press duplicate 3 times. You end up with a total queue of 12 files. The file names are irrelevant, but adhere strictly to the instructions in the Readme file as described in step 2. Run this queue.
When the queue is finished, exit AME. Go to Explorer, navigate to
C:\Users\<user name>\AppData\Roaming\Adobe\Adobe Media Encoder\5.0 and send the file AMEEncodingLog.txt and Adobe Media Encoder Log.txt to me by PM or attach them here and send them to email@example.com.
I will withhold publication till I have seen those two files.
You may have downloaded before this warning was added:
Your data, although complete, raise too many questions. Hence my request.
Thanks for your feedback. I reran the basic benchmark more carefully and found I failed to close Premiere before running AME and this has a huge affect on results, interesting. Attached are the output files from this run. Another change is that I was able to find a stable OC at 3.4. Sorry about the mistake and strange results.
I have not yet run the additional tests you requested seeing how the new run may have resolved the accuracy issues of my first results. I would be happy to run additional tests is you think they are needed to get an accurate measure.
Any suggestions to draw more performance from my system?
My emails to the firstname.lastname@example.org address were returned as undeliverable and PM has not file attachment option.
Thanks for the update. That is much more believable. I forgot to mention one thing, which shows in your render results. It does not matter, since your original data were quite normal as far as rendering. I have included your data, but would like to get both AME text files as well. It may shed some light on what happened the first time.
The reason your emails did not get to me was the extra "5" in the address. The address is ppbm (at) att.net.
Thanks for following up and providing corrected data and a clue to what may cause others may see if they do the same thing.
Harm, I looked in the directory path you instructed, but my system has no appdata folder.
My MPE GPU acceleration was only 4.1X. I recently learned that the PCIe slots on my EVGA P55 micro are both limited to 8x rather than 16x speed. Could this be the reason for the low GPU performance gain?
Any news with the GTX460 card?
Does the Mercury Playback Engine work with this Card?
I started with a GTX260 on my i7 rig last year, then went to a GTX285 this past spring to gain MPE support. Two weeks ago I sold the 285 and got the GTX470. Wow. Best $299 I've spent. I might suggest that since you've come this far with your rig, why monkey around to save $100? The performance gain from the 470 is phenominal, and even though I got it going with the text file hack, the 470 became a supported card with the update Adobe pushed out the other day. So consider bagging your lunch for a while, bypassing the 460, and going for the 470, it won't dissapoint. I went to EVGA directly and they had the SuperClocked edition for $324.99 with a $25 rebate, better price than any online retailer.
And before you balk at the rebate, EVGA is the fastest rebate processor on earth. I mailed mine on a Thursday two weeks ago. The following Monday I got an email saying "Received". Wednesday I got the "Rebate Processed" email, and Friday I got the "Rebate Shipped" email. should be here this week. Seriously, they are that quick. I'm just sayin'.
The problem is that MPE is still in its infancy and the use of all the cores is not yet optimized. The consequence is that it is still difficult to see clear performance gains from better specced cards, but it seems reasonable to expect that with each new update that difference will become much clearer.
No Bill, I can't prove it, I'm just going off of how my rig responds now. However, your question made me realize that my pitch for the 470 is really based on my specific world, not JUST Premiere. I do see gains in Pr but that's cutting together my native 5D Mark II footage, I don't work with any other format. But I realize that the 5D2 native files play much smoother in QuickTime now with the 470, so it's entirely possible that the standard gain in Pr with the 470 is NOT much more than a 285, especially with other formats. For the strange flavor of H.264 QT files that the Canon spits out, the 470 is handling it all quite well. I guess my "phenomenal" description was more my overall user experience, not just Premiere. Windows Benchmark rates the 470 at 7.7 FWIW, and I literally got a 300% frame rate increase with Flight Simulator X with the 470, which I realize has zero to do with Premiere.
However, I still think the 470 is the wisest choice to buy right now, because, like Harm said, Adobe isn't leveraging all that these cards can do. But when Adobe gets out of infancy stage I believe the 470 SC powerplant is going to be way more useful than what the 460 or the 285 offers. Flight Sim and Quicktime proved that to me (yes, the 480 is the big honcho, but it comes with a price, so did the i7 965). Ultimately, I found it interesting that the only 400 series card that Adobe officially supports now is the 470, so it makes me wonder if they will expand their research and capability on that card first before the others. Or, it could just be that the 470 was easiest to test and add in, lol!
Paulie, sure wish I could test a few more but that is the best I can afford to do.
P.S. Harm, the script works wonderfully!
Completely understand that! The 480 is a stretch for me. BW, on this chart you posted, I'm poco un ignaro (italian for a little ignorant), what's better larger or smaller numbers? Just curious...