NVIDIA GeForce GTX 780 Ti Reviewed

[Guest_Jim_*]

I thoroughly understand what officially supporting something would entail, which is why I made the statement that not supporting something is different from attempting to stifle it. The latter is what nVidia is doing and many people would be quite satisfied with the hybrid system simply not being supported, but allowed to work. For crying out loud, overclocking is not officially supported either, but it is allowed. If it were officially supported, then there would be no warnings of clauses about voiding warranties.

An AMD card not being able to run PhysX may be AMD's fault, but a hybrid system being artificially prevented from running PhysX is squarely nVidia's fault. Again, not supporting something and actively trying to prevent something are significantly different.

[Waco]

2. Now as for supporting hybrid builds, you don't seem to understand what officially supporting something would entail. I'm sure most people on this site have come to realize there's a difference between official support and something working. Official support means things have actually been tested and if problems occur, you can issue a support/bug ticket and there's an expectation the issue will be addressed. While it's all well and good that an AMD+NVIDIA-PhysX setup works, if it was officially supported by NVIDIA, they would have to ensure it worked with all sorts of configurations.

 

 

There's a big difference between not officially supporting something and blocking something. Nvidia did the latter. That took time away from useful development to block versus simply stating that it was an unsupported configuration.

[d6bmg]

 

Performance of this card is awesome, but I am a bit skeptical about its price compared to 290x & 280x.

Although I think price will come down in time..

NVIDIA cards do offer several features that AMD cards do not, which is why I think we'll continue to see a small price premium indefinitely.

 

 

Yes it does, a little and the drivers for SLI is more matured than that of CFX, but the question is, when comparing the performence of single card, are those extra features really worth the extra money?

For me, no.

[VaporX]

ClayMeow, on 15 Nov 2013 - 9:59 PM, said:

1. NVIDIA offered AMD PhysX support. That's a fact. The reason why AMD rejected it is irrelevant. Whether they felt OpenCL was the better physics platform or not doesn't matter. What does matter is that NVIDIA pushed PhysX hard and many developers, engines, and games support it. I never said AMD was bad for going another route, I simply stated that PhysX is something NVIDIA offers that AMD doesn't.

 

2. Now as for supporting hybrid builds, you don't seem to understand what officially supporting something would entail. I'm sure most people on this site have come to realize there's a difference between official support and something working. Official support means things have actually been tested and if problems occur, you can issue a support/bug ticket and there's an expectation the issue will be addressed. While it's all well and good that an AMD+NVIDIA-PhysX setup works, if it was officially supported by NVIDIA, they would have to ensure it worked with all sorts of configurations.

 

For example, let's make believe official support existed and NVIDIA releases new drivers. Let's take a recent game like BAO. What if after the new drivers, PhysX worked for everyone except for those that were using a 290X + any NVIDIA card. Now what? With official support from NVIDIA and not AMD, NVIDIA would be responsible for fixing that in the eyes of the consumers. In other words, NVIDIA engineers should have test systems with every AMD card that comes out just to ensure official support. Yeah, that seems logical.

 

This is not a knock on AMD. I'm not saying AMD is bad because they chose to go a different route. What I'm saying is those calling NVIDIA the bad guys because PhysX is exclusive is in the wrong. And if you HAVE to put blame on someone, put it on AMD, not NVIDIA.

 

 Really need some fact clarification here...

 

PhysX was NEVER offered to AMD. PhysX is nothing but a CUDA based program and to offer it you must give CUDA as well, something NVidia would never do. The myth this surrounds this has been convoluted badly. In 2008 a group at NGOHQ made a so called hybrid drivers that was supposed to have allowed CUDA to run on AMD based cards and thus supposedly PhysX. This team approached AMD and asked about being given a number of cards, support for driver development and so on to make this work. At the same time NVidia was contacted to help as well. After waiting a fair period to see the NVidia position and not ever getting any response AMD came to the decision that NVidia would not actively support the project or would find some way around it if they did and so rejected the cost as effort as to not be worth the results.

 

A few days AFTER the AMD decision was made public NVidia, which up until then was silent on this came out and said they would have happily supported the project if AMD had. While a myth has grown from this that NVidia wanted to work with AMD this is purely false. This was a pure marketing attack that allowed NVidia to appear to have taken the high ground in this matter while never having to do anything but keep their mouth shut. A second attempt at making this work BTW came with the same result, NVidia refused to commit even if AMD did commit, this time AMD stayed silent.

 

As for supporting a hybrid build, NVidia does "officially" support a Hybrid build, the official support is to stop it. They do not do anything passive with this but actively put code into their drivers specifically looking for AMD products of video in the system and thus stopping PhysX from functioning. In effect crippling their own product when the competition is present, reducing the ability of the consumer to use the product the way they choose.

 

Finally as for current PhysX support, NVidia likes to say it is open, the reason is they do allow the CPU based version as open. Of course that version is poorly optimized, barely has an effect and is way inferior to the GPU version. This BTW the is the version offered for the XBONE and PS4. So the offer of support for these devices, while a nice press release, is actually not new as the CPU version would work with any current CPU.

[IVIYTH0S]

 

Finally as for current PhysX support, NVidia likes to say it is open, the reason is they do allow the CPU based version as open. Of course that version is poorly optimized, barely has an effect and is way inferior to the GPU version. This BTW the is the version offered for the XBONE and PS4. So the offer of support for these devices, while a nice press release, is actually not new as the CPU version would work with any current CPU.

 

I still don't know why this is, why can't our greatly underused CPUs have the same or nearly the same PhysX experience, they are so underworked by today's games. Is the process REALLY that much easier on a parallel computing device like a graphics card that a powerful multi purpose CPU can't at least meet the demand halfway.

[VaporX]

IVIYTH0S, on 16 Nov 2013 - 3:21 PM, said:

 

 

Finally as for current PhysX support, NVidia likes to say it is open, the reason is they do allow the CPU based version as open. Of course that version is poorly optimized, barely has an effect and is way inferior to the GPU version. This BTW the is the version offered for the XBONE and PS4. So the offer of support for these devices, while a nice press release, is actually not new as the CPU version would work with any current CPU.

 

I still don't know why this is, why can't our greatly underused CPUs have the same or nearly the same PhysX experience, they are so underworked by today's games. Is the process REALLY that much easier on a parallel computing device like a graphics card that a powerful multi purpose CPU can't at least meet the demand halfway.

 

This is a great question and the answer is yes. Physic in and of itself is actually a very simple set of calculations but to work the way PhysX and other implementations do you are in essence doing that simple calculations a thousand times for each movement. Think of an explosion, all those particles are not a single calculation but need to each individually be tracked and calculated. THIS is the strength of a parallel processor, the ability to do all this simple work at once instead of time sharing the load. The CPU has plenty of horse power for the work but not the efficiency in design to split the load as it should.

[Guest_Jim_*]

My way of thinking of it is to compare GPUs and CPUs to calculators. CPU cores are like graphing calculators that can perform just about any mathematical function you want. GPU cores are more like scientific calculators that can really only do the geometric operations. While a CPU may have a handful of cores, a GPU can have hundreds or thousands. Physics calculations are largely linear algebra, which uses the geometric operations, but there can be a lot of calculations to perform. That gives the GPUs a decided advantage because the advanced functions a CPU can perform are just not needed. Each physics particle can have its own little core in a GPU, its own calculator to figure out how it will interact with the environment and each other. Not so in a CPU

 

Now, maybe through some pre-computing schemes a CPU could match the performance of a GPU for calculating physical interactions, but otherwise, yeah, that parallel computing is pretty important. I doubt we'll see that much though, unless the developers look to it to improve a console's capabilities.

[IVIYTH0S]

Then shouldn't APUs be able to fair decently (I'm also including newer intel cpus with the iGPUs)??

[Guest_Jim_*]

Theoretically, yes, but that will require using OpenCL, DirectCompute, or some other standard AMD and/or Intel have access to. Of course an APU is less powerful than a GPU, but it would still be better than a traditional CPU.

[IVIYTH0S]

Theoretically, yes, but that will require using OpenCL, DirectCompute, or some other standard AMD and/or Intel have access to. Of course an APU is less powerful than a GPU, but it would still be better than a traditional CPU.

Precisely!

[Bosco]

 

ClayMeow, on 15 Nov 2013 - 9:59 PM, said:

1. NVIDIA offered AMD PhysX support. That's a fact. The reason why AMD rejected it is irrelevant. Whether they felt OpenCL was the better physics platform or not doesn't matter. What does matter is that NVIDIA pushed PhysX hard and many developers, engines, and games support it. I never said AMD was bad for going another route, I simply stated that PhysX is something NVIDIA offers that AMD doesn't.

 

2. Now as for supporting hybrid builds, you don't seem to understand what officially supporting something would entail. I'm sure most people on this site have come to realize there's a difference between official support and something working. Official support means things have actually been tested and if problems occur, you can issue a support/bug ticket and there's an expectation the issue will be addressed. While it's all well and good that an AMD+NVIDIA-PhysX setup works, if it was officially supported by NVIDIA, they would have to ensure it worked with all sorts of configurations.

 

For example, let's make believe official support existed and NVIDIA releases new drivers. Let's take a recent game like BAO. What if after the new drivers, PhysX worked for everyone except for those that were using a 290X + any NVIDIA card. Now what? With official support from NVIDIA and not AMD, NVIDIA would be responsible for fixing that in the eyes of the consumers. In other words, NVIDIA engineers should have test systems with every AMD card that comes out just to ensure official support. Yeah, that seems logical.

 

This is not a knock on AMD. I'm not saying AMD is bad because they chose to go a different route. What I'm saying is those calling NVIDIA the bad guys because PhysX is exclusive is in the wrong. And if you HAVE to put blame on someone, put it on AMD, not NVIDIA.

 

 Really need some fact clarification here...

 

PhysX was NEVER offered to AMD. PhysX is nothing but a CUDA based program and to offer it you must give CUDA as well, something NVidia would never do. The myth this surrounds this has been convoluted badly. In 2008 a group at NGOHQ made a so called hybrid drivers that was supposed to have allowed CUDA to run on AMD based cards and thus supposedly PhysX. This team approached AMD and asked about being given a number of cards, support for driver development and so on to make this work. At the same time NVidia was contacted to help as well. After waiting a fair period to see the NVidia position and not ever getting any response AMD came to the decision that NVidia would not actively support the project or would find some way around it if they did and so rejected the cost as effort as to not be worth the results.

 

 

 

That's interesting because at a past CES press Conference AMD claimed the opposite. They went on and on how they did not like PhysX and it wasn't open source and they refused to partner with NVIDIA over it. So was the AMD crew giving this speech full of crap or is there more too it? I never thought for once NVIDIA made the offer, however if one was made I could see it being for money to be allowed to use it.

 

As for you guys saying NVIDIA blocked AMD on PhysX blah blah blah. Who gives a shit really? Its NVIDIA's software, they own it they can decide who uses it and who doesn't. We have been over this a million times, and anyone that saw what PhysX could do was amazed. The very first Batman was the first popular title that made it look good, and made AMD look like crap, and yes it was that bad no denying it. NVIDIA had a lot more cash at the time and was able to break ground with Cuda where as AMD was strapped.

 

Both Companies have moved on and are coming out with new things for us consumers. However claiming Mantel and TrueAudio are great things and are going to be so great is a laughable statement

 

AMD hasn't proven anything when it comes to these two so called big things. All we are getting is promises, until its out and tested it really doesn't exist at this point now does it? However, I am all for it helping in anyway, but the little blurb where they say that they are working on getting developers to sign up for TrueAudio is concerning. You can read a lot of things into the statement, but I for one don't expect TrueAudio to pan out much, without developers on board it will not work for that title you are playing based on what AMD has stated. If that's correct, given the small amount of titles that have support, they are going to have a big uphill battle. I for one will stick with EAX its supported a lot more then TrueAudio is.

 

I have to laugh when I read this when someone stated that TrueAudio is the PhysX of Audio, you have to admit its kind of funny

[Guest_Jim_*]

I'm curious about the timing of that CES conference, because if that came after the events Ed described, then it could have been an attempt of AMD's to save face by simply saying they had no interest after the fact. By the way, it is worth noting that as I looked up the offer Andrew mentioned I found that apparently nVidia originally stated that PhysX would be an open-API, and it was in that way that AMD would have had access to it. About a year later nVidia did their complete about face to restrict access to PhysX. And I give a shit, really, every time Andrew or anyone tries to just say AMD was offered the technology but said no, as though that were the end of the story. It is not the end because of how nVidia has locked down the technology. Of course nVidia has the right to restrict access to a technology they own, but it should not be neglected that they have done so.

 

Also, I suspect that some people would say that not only AMD looked like crap with the release of Arkham Asylum. That's the game that exposed nVidia for blacklisting AMD cards from graphical options they are able to run. In fact, because of how the anti-aliasing was added to the engine, AMD cards were already doing the work, but not showing it. I'm not denying anything, I'm just remembering another part of what happened.

 

I have to admit boss, your stance on Mantle and TrueAudio is confusing. I mean every new technology introduced by either company starts off as promises and is only going to succeed if developers get on board. What is laughable about a company marketing a technology of theirs? It would be weird if they didn't. It may be laughable if the technologies do not meet expectations, but it will be months before that can be truly determined for either Mantle or TrueAudio. Months of both potentially influencing console gaming, which would likely mean it will also influence PC gaming. Laughable just seems like a very strong word for something not out yet.

I will admit though, I doubt TrueAudio is going to mean much to the consumer simply because a lot of people don't care about the audio. If it is easy to implement, compared to other solutions, then developers may like it, but I have too low an opinion of the ears of the average consumer. Still, it could achieve a crescendo. It would just surprise me.