• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Is it worth getting a Low End Geforce for CUDA processing?

W

WAKASHIZuMa

Limp Gawd
Joined
Sep 23, 2009
Messages
193
Hello everyone

So I'm getting an ATI Radeon 5850 as my primary graphics card for gaming. I have noticed however that when it comes GPU applications so the support is rather limited for ATI cards while nVidia seems to be ahead with their CUDA programming. I do a lot of video processing on my 920 system and I was curious if there is of any benefit of getting a low end nVidia card (GT 240) for the sole purpose of CUDA processing in the applications that support it

PS: Does Windows 7 support running 2 different graphics cards?
 
If you're thinking about games like Just Cause 2 then no, you can't use one card for DirectX and another for CUDA. If you're talking about running CUDA apps separately then sure but what apps are you thinking of?
 
I believe he wants to use the CUDA technology in nvidia GPUs to do photoshop and video editing work. CUDA makes rendering and encoding video in adobe premiere, DVDFab or Badaboom 5x faster.
 
I would buy the card that supports yor primary habit, games, ati. Folding, badaboom, etc. nvidia. I wouldnt try running both on one pc, if it were me. Just build more boxes!
 
It would be more time-effective to ask an nVidia customer service rep or post on their forums, imo. Running mixed cards has been a bit of a... well mixed bag.
 
from what I am seeing, the benefit of using GPU encoding video is fast but it sacrifice quality for it...

so I wouldn't dare to use GPU encode for now.....
 
dccmadams said:
I would buy the card that supports yor primary habit, games, ati. Folding, badaboom, etc. nvidia. I wouldnt try running both on one pc, if it were me. Just build more boxes!
Click to expand...

Why not, this isn't the days of windows 98 or even XP where having 2 cards caused BSOD all day long.
 
I usually get the ATI cards and I think their 5850 is an excellent choice but I'm disappointed by ATI,s
lack of progress on GPU applications. Besides gaming, I want a card that has good support for gpu processing
as well as Video decoding

So what do you guys think?
Get ATI Radeon 5850 + GT240

no nVidia

wait for Fermi cards
Posted via [H] Mobile Device
 
Seeing how Femi is three weeks away, it would probably be worth the wait. Using a low end card for CUDA wouldn't perform as well as a high end card for CUDA.
 
Could be, but the one thing we can pretty much guarantee is that it's CUDA processing would be faster than a 5850+GT240. If you heavily use applications that support CUDA, it makes sense to at least see what is coming out.
 
WAKASHIZuMa said:
I usually get the ATI cards and I think their 5850 is an excellent choice but I'm disappointed by ATI,s
lack of progress on GPU applications. Besides gaming, I want a card that has good support for gpu processing
as well as Video decoding

So what do you guys think?
Get ATI Radeon 5850 + GT240

no nVidia

wait for Fermi cards
Posted via [H] Mobile Device
Click to expand...

As much as I dislike nVidia, if you're one of the new people that will actually use CUDA for processing there really is no choice. Get a 470 or 480 when it comes out.
 
Does ATI Radeon 5850 support natively decoding all the HD video formats (using Windows Media Player) + flash 10.1 or is that something that only nVidia cards do?
 
WAKASHIZuMa said:
Does ATI Radeon 5850 support natively decoding all the HD video formats (using Windows Media Player) + flash 10.1 or is that something that only nVidia cards do?
Click to expand...

The 5850 will do it as well, although if you're running an i7 it doesn't really matter.

WAKASHIZuMa said:
Hello everyone

So I'm getting an ATI Radeon 5850 as my primary graphics card for gaming. I have noticed however that when it comes GPU applications so the support is rather limited for ATI cards while nVidia seems to be ahead with their CUDA programming. I do a lot of video processing on my 920 system and I was curious if there is of any benefit of getting a low end nVidia card (GT 240) for the sole purpose of CUDA processing in the applications that support it

PS: Does Windows 7 support running 2 different graphics cards?
Click to expand...

I doubt it. A low end nvidia card is going to be pretty slow for CUDA work as well - I have a feeling the extra overhead from using GPGPU to encode videos will likely offset any speed advantage the GPU has over your 920.
 
Thank you guys for your help

I decide to get a 5850 without any nVidia cards. I'm not going to wai for Fermi mostly because it will probably be too expensive at launch

Edit = I have a good quality 750W power supply for my system. Is it enough to drive a 5850?
 
WAKASHIZuMa said:
I decide to get a 5850 without any nVidia cards. I'm not going to wai for Fermi mostly because it will probably be too expensive at launch
Click to expand...

Do you actually have fermi prices?? Please share!

or are you just assuming based on rumors?
 
Depends on what your during with it. If you actually use it on a regular basis then yes. Video encoding for example, I use tmpgenc 4.0 because I strongly believe that no other video encoding software out there even comes close to it in any way, but unfortunately it only supports cuda.

I doubt it. A low end nvidia card is going to be pretty slow for CUDA work as well - I have a feeling the extra overhead from using GPGPU to encode videos will likely offset any speed advantage the GPU has over your 920.
Click to expand...

Don't mean to sound offensive with this but go look at some benchmarks, that statement is simple not true. In fact some low end cards actually do better than high end because MOST cuda apps can only use 32 or 64 SP so as long as you have at least that many SP the shader clock is all that really matters (8800gt is an excellent folding card).

http://www.pcgameshardware.com/?menu=browser&article_id=648270&image_id=841300

FOLDING@HOME is much better with cuda than video encoding and even their you don't see much reason to get say a GTX 260 over an 8800GT (260 is only 27% better in folding@home but is twice as fast in games). I would highly recommend you get a 9800GT or used 8800GT over a 240 though. Buying a high end card for cuda isn't a good idea since they don't scale as well as they do in games.

Also almost any low end nvidia gpu provides significant speedup over an I7 920 in a variety of popular video encoding software.
 
NaturalViolence said:
Depends on what your during with it. If you actually use it on a regular basis then yes. Video encoding for example, I use tmpgenc 4.0 because I strongly believe that no other video encoding software out there even comes close to it in any way, but unfortunately it only supports cuda.



Don't mean to sound offensive with this but go look at some benchmarks, that statement is simple not true. In fact some low end cards actually do better than high end because MOST cuda apps can only use 32 or 64 SP so as long as you have at least that many SP the shader clock is all that really matters (8800gt is an excellent folding card).

http://www.pcgameshardware.com/?menu=browser&article_id=648270&image_id=841300

FOLDING@HOME is much better with cuda than video encoding and even their you don't see much reason to get say a GTX 260 over an 8800GT (260 is only 27% better in folding@home but is twice as fast in games). I would highly recommend you get a 9800GT or used 8800GT over a 240 though. Buying a high end card for cuda isn't a good idea since they don't scale as well as they do in games.

Also almost any low end nvidia gpu provides significant speedup over an I7 920 in a variety of popular video encoding software.
Click to expand...

Do you have any benchmarks of video encoding? All I've seen are old ones like this: http://www.anandtech.com/video/showdoc.aspx?i=3339 which show that even a GTX 280 isn't even twice as fast as a Core 2 Quad. Swap out the GTX 280 with a GT 240 and the C2Q with an i7 and I could very easily see the i7 winning.
 
Headfoot said:
Do you actually have fermi prices?? Please share!

or are you just assuming based on rumors?
Click to expand...

Just assuming based on rumors
Besides CUDA and video codec support I see no reason to wait specifically for nVidia GPUs. Now that I know ATI cards also support video decoding as well as flash 10.1 there is no reason to wait for nVidia
Posted via [H] Mobile Device
 
WAKASHIZuMa said:
I usually get the ATI cards and I think their 5850 is an excellent choice but I'm disappointed by ATI,s
lack of progress on GPU applications. Besides gaming, I want a card that has good support for gpu processing
as well as Video decoding

So what do you guys think?
Get ATI Radeon 5850 + GT240

no nVidia

wait for Fermi cards
Posted via [H] Mobile Device
Click to expand...

DX11 direct Compute and OpenCL both supported not just by ATI.
They will work on every DX10 GPU that is out there.

Im not going to google for you but basically there were lots of announcement of programs using directcompute for video encoding applications when Windows 7 was launched.
 
Pieter3dnow said:
DX11 direct Compute and OpenCL both supported not just by ATI.
They will work on every DX10 GPU that is out there.

Im not going to google for you but basically there were lots of announcement of programs using directcompute for video encoding applications when Windows 7 was launched.
Click to expand...

Thank you

Is Directcompute sort of a competitor to Open CL?
Posted via [H] Mobile Device
 
WAKASHIZuMa said:
Is Directcompute sort of a competitor to Open CL?
Click to expand...

Yup. DirectCompute is Microsoft's answer to OpenCL.

NaturalViolence said:
Directcompute is an api, opencl is a language.
Click to expand...

Both are APIs and both are languages, sort of. DirectCompute leverages HLSL shaders whereas OpenCL uses something new-ish. Ultimately OpenCL, CUDA, and DirectCompute all tackle the same problem with only slight differences in their approaches.
 
WAKASHIZuMa said:
I usually get the ATI cards and I think their 5850 is an excellent choice but I'm disappointed by ATI,s
lack of progress on GPU applications. Besides gaming, I want a card that has good support for gpu processing
as well as Video decoding

So what do you guys think?
Get ATI Radeon 5850 + GT240

no nVidia

wait for Fermi cards
Posted via [H] Mobile Device
Click to expand...

are you saying "encoding"?
You might want to reasearch more before you make the claims, there are some video transcoding are as good if not better than the nvidia part.
 
Both are APIs and both are languages, sort of. DirectCompute leverages HLSL shaders whereas OpenCL uses something new-ish. Ultimately OpenCL, CUDA, and DirectCompute all tackle the same problem with only slight differences in their approaches.
Click to expand...

Hmmm, I was reading the wikipedia article on opencl a few days ago and it specifically states that although opencl has it's own api opencl the language will run in cuda and directcompute as well. And I know that directcompute exposes the shader pipeline to an application the same way cuda does. Are you sure they are rivals? I mean I could be wrong but I am fairly certain that both cuda and directcompute run opencl code.

Edit: http://www.nvidia.com/object/cuda_directcompute.html

I knew it. Directcompute can run opencl, cuda c, fortran, and c. CUDA can run cuda c, opencl, c, and c++. Opencl (the api) can only run opencl (the language). Directcompute has no single prioritized language named after it.
 
Last edited:
NaturalViolence said:
Hmmm, I was reading the wikipedia article on opencl a few days ago and it specifically states that although opencl has it's own api opencl the language will run in cuda and directcompute as well. And I know that directcompute exposes the shader pipeline to an application the same way cuda does. Are you sure they are rivals? I mean I could be wrong but I am fairly certain that both cuda and directcompute run opencl code.

Edit: http://www.nvidia.com/object/cuda_directcompute.html

I knew it. Directcompute can run opencl, cuda c, fortran, and c. CUDA can run cuda c, opencl, c, and c++. Opencl (the api) can only run opencl (the language). Directcompute has no single prioritized language named after it.
Click to expand...

That link doesn't say anything like what you are claiming. Nvidia decided to implement OpenCL and DirectCompute by wrapping them in CUDA. CUDA can only run cuda c, that is the only language it supports. Neither CUDA nor DirectCompute can run OpenCL. OpenCL is its own thing.

You are confusing how Nvidia is implementing OpenCL and DirectCompute with how developers call and use them. Two *very* different things.
 
CUDA is nvidias proprietary answer to gpgpu, OpenCL and DirectCompute are open for ANYBODY to support and use,
Stream is ATI/AMD's proprietary method, BUT its sort of being ignored due to OpenCL and DirectCompute being a preferable option since they arent limited to one hardware manufacturer, they could also both be run on cards and chips that have OpenCL/DC drivers, an cell processor card like that one toshiba was working on could run OCL/DC software so could a card that had an x86 chip like the nano or atom on it, exciting stuff, we could see accelerator cards start to appear again that could not only accelerate games(like the voodoo cards did) but could be used to accelerate other software without slowing the rest of the system down or being cpu dependant.
 
Stream was being ignored long before OpenCL came on the scene due to its immaturity (compared to CUDA). In terms of OpenCL vs DirectCompute they aren't really competitors. The latter is meant to be used in games in conjunction with DirectX. OpenCL is targeted at more general applications.
 
You must log in or register to reply here.
Back
Top