1.
show the post

Discussion
Discussion on Article:
Do Graphics Cards Need 4 GB of Memory? - EVGA GeForce GTX 670 4 GB Superclocked+ Review
Started by: MyBrainIsMelting | Date 08/07/12 12:49:27 AM
Comments: 11 | Last Comment: 11/01/12 05:26:27 PM
Expand all threads | Collapse all threads
[1-7]
2.
It would be nice to run the same tests with (at least) 2-way sli.
And then, run the whole thing again with 2560x1440@120Hz, surround/infinity of 2560x1440 @60Hz and then @100Hz (since nVidia drivers are bogus in sli).
Finding such a monitor is not easy, though, so I guess sticking to 60Hz at first would be ok.
And then, run the whole thing again with 2560x1440@120Hz, surround/infinity of 2560x1440 @60Hz and then @100Hz (since nVidia drivers are bogus in sli).
Finding such a monitor is not easy, though, so I guess sticking to 60Hz at first would be ok.
3.
About your test with Sniper Elite V2 Benchmark:
"This is the most resource-consuming application on our list. It needs 1.3 gigabytes even at 2560x1440 without antialiasing and 2 gigabytes with 4x MSAA."
Nope, this bench doesn´t use Multisampling antialiasing, it uses SUPERsampling (in the file "settings.ini" you can config the test, you can use supersampling, 2.25x or 4x, and not to use it).
The tremendous waste of memory with antialiasing in this test is related with the supersampling technique. With 2560x1440 and 4x, you are using "internally" 5120x2880 (56 MB in one framebuffer with 32 bits of color) and downsampling it.
"This is the most resource-consuming application on our list. It needs 1.3 gigabytes even at 2560x1440 without antialiasing and 2 gigabytes with 4x MSAA."
Nope, this bench doesn´t use Multisampling antialiasing, it uses SUPERsampling (in the file "settings.ini" you can config the test, you can use supersampling, 2.25x or 4x, and not to use it).
The tremendous waste of memory with antialiasing in this test is related with the supersampling technique. With 2560x1440 and 4x, you are using "internally" 5120x2880 (56 MB in one framebuffer with 32 bits of color) and downsampling it.
4.
This card is a waste. 4GB are indeed god for multi-monitor gaming, but then again, this card is to slow anyways for that. Would be nice if those 2 GPU monster cards would have 4GB or more for each GPU
5.
Yes, they do need.
And not for the video games, but for GPGPU.
some here will say you have nvidia tesla. yes we do have it, but it cot several thousand euros, while geforce is 500 or 600 euros.
And not for the video games, but for GPGPU.
some here will say you have nvidia tesla. yes we do have it, but it cot several thousand euros, while geforce is 500 or 600 euros.
6.
Please add CUDA code testing. These test for games are not interesting. I suggest you to write a CUDA code that will use 4 GB of memory or more and will run for several hours. These results would be interesting. Also CUDA on dual gpu or triple. Thanks in advance.
7.
Are you Bad it's because cuda aplications, for example I work with Linux an virtualize all time, and play hard to, I really need prosesors with 8 cores efectives an a lot of grafic memory
[1-7]



| Date: 08/07/12 12:49:27 AM]
| Date: 08/07/12 06:06:38 AM]
| Date: 08/09/12 09:09:43 AM]
| Date: 08/10/12 06:35:48 AM]
| Date: 08/22/12 03:20:49 AM]

