GeForce 7200 GS

[MajorCaliber]

I have an unused desktop with an old processor that will not produce much work. It has a dead graphics card and I have a brand new in the box GeForce 7200 GS. Is this card capable of FAH, and any idea of the PPD it might do? I'm trying to get an idea if it's worth my time to mess with it.I know very little about graphics cards because I'm not a gamer.

[Joe_H]

The Geforce 7200 GS is a couple generations too old to be used for Folding@home. About the oldest cards that have partial support are the GT 400 series cards, and in practice some of the GT 600 series are about the oldest that can be used.

[JimboPalmer]

The Geforce 7 series used hard wired vertex shaders and texture shaders. They cannot be reprogrammed to do folding.

The Geforce 8 series was the first to used re-programmable unified shaders. As Joe_H explains, that generation is now too wimpy to do F@H, but it has the correct idea. (It can't do 64 bit double precision Floating Point math, and has no OpenCL 1.2 software support, which F@H now requires)

https://en.wikipedia.org/wiki/GeForce_8_series

[MajorCaliber]

Thanks for the advice. I suspected it was too old, that's why I checked before I spent the time to try it.

[Joe_H]

If you come across other older cards, new or used, for nVidia based ones you would be wanting to use those based on their Kepler or newer GPU chips. The Kepler chip was introduced starting in the GTX 600 series cards. For F@h the Kepler based cards are the oldest for which use of CUDA instead of OpenCL instructions for the GPU folding core is supported.

The previous generation Fermi chips used in the GT 400, 500, and some lower end 600 or 700 series cards only fully support OpenCL 1.1 in hardware. F@h GPU folding was coded to the OpenCL 1.2 standard. Some of the OpenCL commands introduced between the 1.1 and 1.2 standards are available on the Fermi chips, which instructions are present varies depending on the specific chip. The GPUs can work successfully on some projects and fail on others depending on the exact instructions needed. I suspect at some point in the future Fermi support will be ended for F@h.

[psaam0001]

I believe the GT 710's I'm running are Kepler based if I am reading the GPU Code name right (GK208B).

So I'm good until at least OpenMM_23/OpenMM_24.

Paul

[Joe_H]

That is correct, GKnnn would be Kepler chips. All but a few GT 700 series cards are based on Kepler or Maxwell chips. The partial list of cards from that series I have looked at has a GT 705 and one variant of the GT 730 which are based on Fermi (GF1nn).

[psaam0001]

Ending support for the Fermi GPU's may also reduce the complexity of the code that powers the F@H client, where it only has to check for CUDA or OpenCL 1.2 support when it receives a GPU WU.

Paul

[bruce]

FAH doesn't check directly for OpenCL1.2 support. That information is inherent in the decisions made based on the information in GPUs.txt. It simply queries the device ID codes and looks it up in that table. It's only when we're presented with DeviceID codes for a newly manufactured GPU that we need to verify if it can do FP64 instructions using OpenCL1.2. It's very rare that we need to do that.

From your perspective, it's easier for you to search the characteristics of the GPU and understand why it can't be supported by FAH than for you to look in GPUs.txt and decode the Species assigned.