My pc build advice/opinions needed

GPUdirect applies to CUDA only...its entirely irrelevant for gaming (kinda hilarious, this being mygaming.co.za). Don't believe me?...go to Nvidia's dev page & see how many references to CUDA you can find. Here:

https://developer.nvidia.com/gpudirect

Besides, DMA has existed for a billion years already, so the notion that one company could cause an implementation of it to be "NOT AVAILABLE TO THE PUBLIC" is somewhat absurd.

Oh and DMA is enabled by default on ATI GPUs...so default in fact that you need 3rd party tools like ATI Tray Tools to disable it.: (Screenshot, my PC)

View attachment 4127

Hell I can even point you to the OpenGL function that *implements* DMA...on both ATI and Nvidia cards. Here is a code example...in the _official_ fckin OpenGL specs (ctrl-f DMA). Now OpenGL as you'd imagine is OPEN. Contrast that with your claim of "NOT AVAILABLE TO THE PUBLIC".

So your "GPUdirect is the only gpu dma technology" theory is looking kinda shaky.


"Calculations"...right...I'll comment when you narrow it down a bit more.


Well I've provided a screenshots, links and an actual fckin code example...you've provided absolutely nothing. Actually no, thats not right, you provided a link to a wordpress article that says "Current GPUs contain a DMA controller". I wonder why it doesn't say "only current Nvidia GPUs". Also wordpress article...not the best of sources. Code examples & official specs definitely ftw. I'd recommend hoisting a white flag.

And finally, there is a limit to:

Ai, ai, ai. I was actually very reluctant to reply , but since I don't want people to believe the absolute donkeycrap posts you've made, here goes :

The GPU has a series of registers that the BIOS maps. These permit the CPU to access the GPU's memory and instruct the GPU to perform operations. The CPU plugs values into those registers to map some of the GPU's memory so that the CPU can access it. Then it loads instructions into that memory. It then writes a value to a register that tells the GPU to execute the instructions the CPU loaded into its memory.

The information consists of the software that the GPU needs to run. This software is bundled with the driver and then the driver handles the responsibility split between the CPU and GPU (by running portions of its code on both devices).

The driver then manages a series of windows into GPU memory that the CPU can read from and write to. Generally, the access pattern involves the CPU writing instructions or information into mapped GPU memory and then instructing the GPU, through a register, to execute those instruction or process that information. The information includes shader logic, textures, and so on.

Regarding the GPUdirect subject : http://en.community.dell.com/techcenter/high-performance-computing/w/wiki/2343.aspx

Do not fucking tell me I am the one who was confused. How can you even say CUDA has nothing to do with gaming, http://www.youtube.com/watch?v=-6t2rrn4hz8 .

I have been going with general knowledge this far , but damn you for making me do research.

On another note, lets take the fighting to the pm if you are still keen to argue. This is supposed to be a fun forum, so lets not ruin the atmosphere , okay?

Peace.

@ JP , you have to know your stuff to be able to argue mate. I have been involved with IT since I was like 10 or 11, but have no official qualifications. I don't know if Havoc works for AMD or Nvidia or is a hardware engineer (just an example), but hell if he tells me right now he has a qualification in any of that I will apologize , admit that I'm wrong and never post about this subject again. If not, we are just two enthusiasts arguing, and won't end until one admits he is wrong.
 
Last edited:
I don't know if Havoc works for AMD or Nvidia or is a hardware engineer (just an example), but hell if he tells me right now he has a qualification
Definitely no an 3D engineer or anything even close. Did a bit of amateur 3D programming (OpenGL & DirectX) but thats about it.:p

The GPU has a series of registers that the BIOS maps. These permit the CPU to access the GPU's memory and instruct the GPU to perform operations. The CPU plugs [...]
While I don't agree with all of that, I can now see where we are talking past each other. You seem to be talking primarily about the instructions. Those I will gladly concede go via the CPU in roughly the manner you described.

I was focusing more on textures since 99.999% of the data the GPU is going to work with will be textures. This is also what matters when it comes to bottle-necking. These textures are moved via DMA.

You are looking at the _high performance computing_ section of Dell's website. What they mean by that is number crunching. i.e. Racks upon racks of GPUs crunching a mountain of numbers using either OpenCL, or your personal favorite: CUDA. The issue with this is that you need to link these GPUs somehow hence the article states:
One of the limitations of the architecture is that GPUs cannot directly perform IO to storage devices or communicate with network devices.
That last part is what GPUdirect fixes. Very cool, but not gaming (or even desktop PC) related.

Do not fucking tell me I am the one who was confused. How can you even say CUDA has nothing to do with gaming, http://www.youtube.com/watch?v=-6t2rrn4hz8 .
As described above, CUDA is intended for crunching numbers. That might sound a bit vague so I'll provide some practical examples that you & everyone else can relate to:
1) Folding@Home uses CUDA to crunch medical simulations
2) Bitcoin miners use CUDA to accelerate their crypto code, making GPU mining much faster than CPU mining

Perhaps there is some hypothetical way to use CUDA for rendering 3D games, but I'm not aware of any. Its neither designed nor (afaik) used for it. Maybe you could acceleration non-rendering aspects of games (pathfinding, AI etc) but dev's tend to just give it a miss entirely for the same reason a PhysX: Only half the gear out there supports it so as a dev you can't rely on it being there...meaning you need to write your code assuming that its not...so why bother with it in the first place. 3D gaming uses standard DirectX/OpenGL calls, plus shader programming (which is somewhat related to CUDA).

Notice how the guy in the second link talks about "software" not "games". There are graphic application where CUDA is used. e.g. Photoshop. Pretty sure its also used for ray tracing algos, but too lazy to check.

I have been going with general knowledge this far , but damn you for making me do research.
Perhaps you learned something during your research. I sure did during mine. :)

On another note, lets take the fighting to the pm if you are still keen to argue. This is supposed to be a fun forum, so lets not ruin the atmosphere , okay?
I don't think its necessary. You just caught me on a bad day - if you feel I was unnecessarily harsh then I apologize for that. I hope the tone in this post comes across a bit friendlier.

the absolute donkeycrap posts you've made
:( /sad
 
build complete... for now

Okay so it's done, at least for now that is.
Here are the specs of how the final build turned out.

Asus P8Z77-M
Intel Core i7 3770K 3.5GHz
Corsair Vengeance 8GB DDR3 1600MHz
Seagate Barracuda 1TB HDD
Asus Eah 4870 512mb DDR5 gpu
Cooler Master eXtreme Power 650w psu
Cooler Master 690 2 plus
Asus DVD writer as i needed a SATA interface

Cost was R7300

Upgrades will hopefully just be a gpu, something like the ATI 7850?? or something from Nvidia?? still need to do some research to find out what I'll go for.

Thats it for now though. Thanks for all the input on this thread
 
Definitely no an 3D engineer or anything even close. Did a bit of amateur 3D programming (OpenGL & DirectX) but thats about it.:p


While I don't agree with all of that, I can now see where we are talking past each other. You seem to be talking primarily about the instructions. Those I will gladly concede go via the CPU in roughly the manner you described.

I was focusing more on textures since 99.999% of the data the GPU is going to work with will be textures. This is also what matters when it comes to bottle-necking. These textures are moved via DMA.


You are looking at the _high performance computing_ section of Dell's website. What they mean by that is number crunching. i.e. Racks upon racks of GPUs crunching a mountain of numbers using either OpenCL, or your personal favorite: CUDA. The issue with this is that you need to link these GPUs somehow hence the article states:

That last part is what GPUdirect fixes. Very cool, but not gaming (or even desktop PC) related.


As described above, CUDA is intended for crunching numbers. That might sound a bit vague so I'll provide some practical examples that you & everyone else can relate to:
1) Folding@Home uses CUDA to crunch medical simulations
2) Bitcoin miners use CUDA to accelerate their crypto code, making GPU mining much faster than CPU mining

Perhaps there is some hypothetical way to use CUDA for rendering 3D games, but I'm not aware of any. Its neither designed nor (afaik) used for it. Maybe you could acceleration non-rendering aspects of games (pathfinding, AI etc) but dev's tend to just give it a miss entirely for the same reason a PhysX: Only half the gear out there supports it so as a dev you can't rely on it being there...meaning you need to write your code assuming that its not...so why bother with it in the first place. 3D gaming uses standard DirectX/OpenGL calls, plus shader programming (which is somewhat related to CUDA).

Notice how the guy in the second link talks about "software" not "games". There are graphic application where CUDA is used. e.g. Photoshop. Pretty sure its also used for ray tracing algos, but too lazy to check.


Perhaps you learned something during your research. I sure did during mine. :)


I don't think its necessary. You just caught me on a bad day - if you feel I was unnecessarily harsh then I apologize for that. I hope the tone in this post comes across a bit friendlier.


:( /sad

Yes I think we have had a big misunderstanding of what we were arguing about, I'm pretty lost by now lol . About CUDA, I'm also pretty lost about that aswell. I thought there were CUDA - enabled games and not just apps, but looks like there are none ( for whatever reason...) . http://hexus.net/tech/news/graphics/47081-cuda-5-kepler-best/ , maybe we see CUDA - enabled games in the future? But still it doesn't make sense for me not to use CUDA as an API in games.
 
+1 in favour of a newer gpu.

I have the Sapphire 4850x2 2gb. Great card, but old is old. Starting to see issues in new games as the card is only dx10. It is also power hungry.

Buying a newer gpu might save you buying a new power supply.
 
Back
Top