Basic card compatible with GeForce GTX 1050

Hi,

Apologies in advance if I’m posing this question in the wrong forum.

I have been upgrading some older workstations with GeForce GTX 1050 cards to run dual 4K screens (all working very nicely).

In one particular case; I needed to keep the incumbent card to drive more monitors. The incumbent card is a GeForce 9500 GT. it was working perfectly until I installed the GTX 1050.

Now, the 9500 GT isn’t recognized in the NVidia control panel. When I looked into the problem, I see an error in the Windows device manager: “Windows has stopped this device because it has reported problems. (Code 43)” I have tried all the usual troubleshooting steps - no joy. The PSU is 850w which is plenty of power for both cards (according to the consumption figures).

I see the GeForce 9500 GT is CUDA based - should it work harmoniously in the same system as the GTX 1050 or is it essentially clashing with it? If so; I’d really appreciate some advice on what basic card I can get that should run alongside the GTX 1050 nicely to run two 1080p monitors - a GeForce GT 710 perhaps?

Any advice greatly appreciated.
Regards
Al

PS: The system spec in use:
Windows 10 Pro x64
ASUS P9X79 MoBO
32GB RAM
i7 3930k
DarkPower Pro 850w PSU

The problem is the 9500 GT is no longer supported by recent drivers nor is it supported by recent CUDA versions. And if you want to use the GTX 950, you will need a recent driver to support it.

There is no solution to this. They cannot both be used in the same system.

As txbob points out, the “incumbent” GT 9500 is hopelessly outdated. I would suggest entrusting it to the care of an approved e-waste recycler of your choice.

The GT 710 appears to be based on the Kepler architecture. This is supported by current drivers and CUDA versions. Given that NVIDIA just discontinued support for the Fermi architecture which preceded Kepler, one could reasonably assume that Kepler architecture GPUs will be supported for a couple more years. If you want something a bit more future proof, you might want to look at the GT 1030 instead, which however is more expensive:

https://www.geforce.com/hardware/desktop-gpus/geforce-gt-1030/specifications

Robert_Crovella, njuffa,

Thanks very much for your detailed replies, much appreciated. I will look to get a GT 1030.

Just out of interest… I looked up the GT 9500 and it seemed to be from 2015… is that incorrect? What the usual lifespan that Nvida support an architecture for please?

Al

I would say that is incorrect. I’m not sure what you “looked up”.

GeForce 9500 GT was a GPU developed in the 2008 timeframe. Here is a press review from 2008:

http://www.legitreviews.com/nvidia-geforce-9500-gt-video-card-preview_760

9500 GT was a member of the GeForce 9 family of GPUs (a derivative of the GeForce 8800 architecture):

https://en.wikipedia.org/wiki/GeForce_9_series

These GPUs were end-of-life (using that word approximately) with the release of the R340 and R343 driver branch, the last drivers that supported that GPU family. Some R340 drivers were still being released by NVIDIA as late as 2016, according to what is available at http://www.nvidia.com/drivers

But compatibility and some definition of usability realistically ended before that. By 2016 there were GPUs released that would not work with R340/R343 drivers. Compute support for these GPUs was dropped in CUDA 7.0, which was released in March of 2015.

If you want to do similar legwork as above, you can come up with a definition of “usual lifespan”. 5 years seems like a reasonable, conservative estimate. Possibly longer. The GPU generation after this one, called “Fermi” was shipping in late 2010, and is currently being phased out (CUDA 9 no longer supports Fermi, and driver support will probably be phased out soon, although R390 still supports Fermi to some degree.)

Interesting question. I am not aware on any recent (i.e. past five years) Geforce model numbers in the 9000 range. I took this to be a G96-based GPU with compute capability 1.1, ca. 2008. At this time, the minimum supported architecture is compute capability 3.0.

The GPU from the 2015 timeframe you might have come cross might have been a GTX 950 (note single trailing zero), based on the Maxwell architecture (GM206, compute capability 5.2).

Thanks again for the swift replies & comprehensive info - I’ve learnt a lot!

I’m not sure where I saw the 9500 GT was a 2015 card… I would have slung it on the recycling heap immediately if I’d released it was 10 years old!

I’m really impressed with the GTX 1050 cards - am I right in saying they’re essentially pretty low in NVidia’s current line-up?

Al

If you consider the total Pascal line-up (GT 1030 to GTX 18080 Ti), it spans roughly one decimal order of magnitude in performance (based on FLOPS and memory bandwidth). If we assign a scale of 1 to 10 to that range, the GTX 1050 rates about a 2.

Notes:

(1) The correct name is apparently GT 1030 (not GTX 1030 as I used earlier)
(2) NVIDIA seems to have released a newer version of the GT 1030 with much slower DDR4 memory