Depends a bit on the rest of the system, but generally the 670 is at least two times faster.
Sandybridge 2600k (3.4ghz quad core)
16GB DDR3 1600
80 GB SSD for OS
60 GB SSD for scratch/cache
2x 2TB 7200rpm data drives
650w coolmaster PS
With this hardware it may be around 2.5 times faster.
Thanks Harm... I'm rendering 1hr of RED footage (2k 12:1) into an H264 (CBR 18Mbps) and its guessing around 7.5 hrs render.
I'll yank this card after the project wraps... put the 4800 FX in and render the same timeline... give me about a week before I post results.
So, finally got the RMA mobo in and just swapped out the mobo and graphics card. CURRENTLY: Quadro 4800 FX and the older ASUS mobo.
The Quadro was faster, barely.
I tested by rendering a 8 second 4k RED clip with some small filters applied...
GTX 670 (with NEW mobo) = 17seconds to render
Quadro 4800 FX (with older mobo) = 16 seconds to render
Used the same sequence and created a preset so they rendered the same output. All other hardwar was the same. Looks like i will be returning about $600 worth of hard ware i did not need.
It sounds like you used very few, if any, CUDA-accelerated effects in your short project. Most of the effects that you used appeared to be non-accelerated (and thus are rendered in software-only mode). Thus, if the CPU has to perform most of the rendering, the Quadros will be faster than GeForces with four to eight times the number of CUDA cores because the GeForce OpenGL drivers are not as good as those in the Quadro drivers. That was completely unexpected (and very disappointing) given that the GTX 670 has 1,344 CUDA cores while the Quadro FX 4800 has only 192 CUDA cores. However, the FX 4800 was a $1,700 card when it was new while the GTX 670 was introduced at $400.
I agree 100% that it is disapointing... I did check to make sure that both graphics cards were useing the Mercury Playback Enging GPU accel (CUDA)... had to mod the txt file for the GTX but it worked.
And yes, the cuda cores were what made me select the 670 big time... Again, this was pretty hefty footage. /shrug maybe Id see a difference if i was doing a 12 camera multicam all RED footage.
On the other hand, the GTX 670 is actually the bigger bang for the buck when comparing their prices when new: The FX 4800 (when new) cost at least four times more money than the GTX 670. I would have been very embarrassed at myself for spending all that money ($1,700) for a GPU that barely outperforms a $200 GPU.
True... if they are a push when it comes to render times and playback... I'd buy the GTX (heck, might even get back into PC Gaming!) but as it stands, I can't return the 4800
I'm late to this thread; just read it for the first time, and I have several observations to share...
1) I tend to agree with Harm and and RjL regarding the GTX being quite a bit stronger; so in your system where this is not evidenced, possibly the video card it is not the "critical path" (or bottleneck) for whatever you are doing (playback, render, etc.). Running Harm and Bill's excellent PPBM5 benchmark can tell you a lot about how the various areas of your system perform and the 800+ system database shows what others are getting from their listed hardware.
2) Premiere Pro is pretty good at taxing all of a PCs resources and "balance" of components is important in order to eek out the best overall performance. I would quickly add another RAID 0 pair to this system if I were you. You can test this theory too by looking at the performance tab in Windows task manager and see if your cpus (all 4 cores and 4 hyperthread cores) are working pretty hard. If they never all running at or very close to 100%, then you have gains to be made by either tweaking the system and/or revising components in your build.
3) You may not even have a motherboard problem; buy or borrow a highy rated 850w or larger power supply and see if that does not clear up whatever woes you are blaming your motherboard on; 650 may not be enough for rendering RED!
I think benchmarks are very relevant when choosing a graphics cards for an editing business situation. But so is performance over time. Failures due to overheating or frugal system design is harder to measure but just as important to evaluate. We are told that the Quadro line is better built and ventilated than the GTX line. It would be nice to see some hard statistics on performance over life in products as well. If a GTX craps out from constant use, its speed is not enough of a good reason to buy it.
I think benchmarks are very relevant when choosing a graphics cards for an editing business situation.
Does posting the same paragraph in more than one thread somehow make it more relevant?
No smartass, just particpating in 2 different conversations. Is that alright with you
hey hey hey... if we are going to get rude target the idiot who knows very little about benchmarking. :ME:
I find it interesting that we are all pretty shocked that the 670 didnt blow the 4800 out of the water, but when I step back, I think, "Should a $400 card really beat a $1,700 one?" (*as RJ pointed out). Granted the 670 is brand spankin new and has more cuda cores. But the 4800 is still a beast.
I've heard the 4800 is crap when it comes to gaming and the flip side, the 670 is one of the top end cards. I've not gamed in YEARS but really am excited for Diablo 3. (*its busy season for me so I still have not picked it up) I downloaded a guest pass before taking my computer apart to do another comparitive 'test'. Of course the 670 chewed through it on the highest settings. <--> Change out mobo/4800 <--> and it played GREAT. I think one setting (details IN the shadows INside the birds ears as they fly by had to be set to high instead of MAX).
The point of my Diablo 3 story is this. Diablo 3 wont require a massive card. Both cards will eat it up. SO maybe What I'm doing is a cake walk for both cards!
((i SHOULD have popped open after effects and really beat up the machine))
It would be interesting to see the PPBM5 Benchmark results on your system, once with the FX 4800 installed, and once with the GTX 670 installed. That will give you a clear answer which card is the better one, performance wise.
If you also go to the Benchmark Results and select from the Video Card panel the FX Series, you will see where the FX series are in the ranking.
I just took a good look at the results list. Of those systems that are running CS5.5 and an FX 4800, none of them are equipped with a current-generation CPU that runs even close to 3.0 GHz (and those systems that run a CPU at even 3.0 GHz use old, pre-Nehalem CPUs). Of those, the fastest one is a system equipped with two last-generation Xeon CPUs running at only 2.4 GHz - and that system ranks 387th out of the current 914 systems on the list, with overall performance that's around 3.7 times slower than a fast overclocked i7 system. The highest-ranked system equipped with an FX 4800 is helped by CS 5.0.3, but still ranks 283rd out of 914.
Actually Harm, I still have a Quadro CX on my shelf which is a 4800 so I will try and get that done for you asap.