Current facts: It is generally available for prices around $ 1000. The specs are impressive, apart from the clock speed.
Current suppositions: My personal view is that it could be 30 - 40% faster than a 680 as I stated in another post, which you obviously read.
Benchmark results: There is only one entry with the Titan in our benchmark and that is not telling us anything, since the conditions under which it was performed differed widely from earlier test results. This one entry has very nice results, but still does not do better on the MPE render test than a GTX 480.
My suggestion: Wait a bit for more news about this card to emerge before you decide to pay this kind of money for a video card.
General advise: Have a look at Where to spend $$$
It is easy to spend a large amount of $$$ on a video card, but when one is wise, the amount spent must be 'balanced' to the total system. IMO the only systems possibly worth a Titan card are overclocked i7 hexa cores or dual Xeon E5-2687+ systems with 64+ GB memory and a very, very nice disk setup with a dedicated raid controller. In all other cases it is overkill and you will not benefit from the expenditure.
Just my $ 0.02
I certainly agree with your advice about waiting. It sure seems to be a possible boon to my computer. which you just about described to a T.
Harm Millaard I'm not going to buy it.. I was just wondering.
But i really want to buy a card in my budget. I checked in PPBM7 GTX 660 has good MPE gain. It fit my budget and also the powersupply. But i didn't see GTX 660Ti on that list. I am wondering may be tha gain retain by the Ti version. For Ti version i don't think my PSU can support. I don't want to overstress it. But how come GTX 660 or Ti had good gain than GTX 670?
The data on the PPBM7 site are pulled from the PPBM5 database, and when I set that database up and made the queries for PPBM5, I had not foreseen this need. So all GTX 660 cards do include a number of GTX 660 Ti cards as well.
That is why we want to use Speccy in the future, to get more detailed info in a consistent way.
Can you answer me?
How does GTX 660 have more gain than GTX 670?
GTX 670 has cores,bandwith better than GTX 660
Gain = (Render time in software mode) / (Render time in hardware mode).
The slower the system, the higher the gain, because the render time in software mode is very long. That is why I said:
CAVEAT: A higher gain value does not by definition mean better. It can be influenced by a slower CPU, single processor and/or slower memory. Average render speed may be a better indicator of performance.
The GTX 670 is more a high-end card than the GTX 660, so it makes sense that you will see the 670 in more high-end systems and the 660 in more middle of the road systems and that is the reason the gain for a 660 can be higher than the gain for a 670. As an example:
Gain Hi-end system, GTX 670 48 seconds 4 seconds 12.0 x Mid-level system, GTX 660 96 seconds 6 seconds 16.0 x
Hope this explains why interpreting 'Gain' is difficult and why average render time may be a better indicator.
Now i understand. But even in render time installed System is a factor. isn't it?
Yes, it is. Remember that rendering in hardware mode depends very much on the speed with which frames can be transferred from regular memory to the GPU to VRAM and back to regular memory. The basic ingredients here are the amount of RAM, number of cores and the clock speed.
Tanks you were so helpfull about the MPE doubts in this forum
I found this article and this make me wonder all cards have simler perfomance when tested in same PC
Not completely true. Bill has tested our new MPEG2-DVD timeline, which is a single AVCHD 1080i/29.97clip with a number of effects like fast color correction, brightness & contrast, 3 way color correction, gaussian blur, gamma correction and speed adjustments and a duration of 2;39;04.
Exporting to MPED2-DVD with the preset NTSC 23.976 Widescreen high quality and MRQ turned on, Bill tested this on his i7-2600K, (OC) 4.4, 32 GB memory and the project on 2 Samsung 840 Pro SSD's in raid0, he found the following results:
Just using two of Bill's observations plus his i7-960X system and my own gives these summary results:
Gain BillG i7-2600K and GTX 680 OC 31 seconds (5.1 x RT) 870 seconds 28.1 x BillG i7-2600K and GTX 680 35 seconds (4.5 x RT) 870 seconds 24.9 x BillG i7-960X and GTX 680 OC ? 30 seconds (5.3 x RT) 556 seconds 18.5 x
Harm's Monster and GTX 680 OC
24 seconds (6.6 x RT) 436 seconds 18.2 x
Here CUDA acceleration is severely tested, since there is scaling, blurring and frame blending going on. Note that the faster a system is, the lower the gain, but also the impact of a faster system on the results in hardware render time.
PS. Matt said it in his introduction, he tested with the PPBM6 file as it existed at that moment and since then it has been changed drastically. In addition Matt used the Total Time score for his comparison where the RPI score would have been better. The Disk I/O test is completely irrelevant when testing hardware acceleration, H.264-BR is largely irrelevant, it mostly is all about the MPEG2-DVD test and - at that moment at least - the rendering test.
Hello Everyone. Yes i tired both cards GTX670 and GTX Titan. What i notice there is no difrence in performance but ya there is difrence in price. I am using 3.5GHZ Intel Quad Core 4th Genreation with 16Gb Ram 2400MHZ.
Since GPU acceleration is a pipeline and the CPU the starting point, the CPU decides how much a GPU will be pushed with data. A Quad Core CPU wont push either card that much especially the Titan.