# GTX 950 + Quadro P400 benchmarks

**URL:** <https://forum.livepeer.org/t/gtx-950-quadro-p400-benchmarks/1497>\
**Category:** Transcoders\
**Created:** [September 17, 2021, 2:45pm UTC](https://forum.livepeer.org/t/gtx-950-quadro-p400-benchmarks/1497 "2021-09-17T14:45:09Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![Strykar](https://yyz2.discourse-cdn.com/flex030/user_avatar/forum.livepeer.org/strykar/32/758_2.png) [@Strykar](https://forum.livepeer.org/u/Strykar)\
**Post date:** [September 17, 2021, 2:45pm UTC](https://forum.livepeer.org/t/gtx-950-quadro-p400-benchmarks/1497/1 "2021-09-17T14:45:09Z")

</div>

Nvidia Quadro P400 (2 GB RAM) (_Pascal GP107 6th Gen NVENC chip_):

```auto
*------------------------------*---------------------*
| Concurrent Sessions | 6 |
| Total Segs Transcoded | 180 |
| Real-Time Segs Transcoded | 180 |
| * Real-Time Segs Ratio * | 1 |
| Total Source Duration | 360.00000000000017s |
| Total Transcoding Duration | 39.94000854000001s |
| * Real-Time Duration Ratio * | 0.1109 |
*------------------------------*---------------------*

```

The P400 will cap out at 6 concurrent streams, it appears limited solely by its VRAM. A hardware mod to swap out for larger RAM chips might work, I could not find anyone who has done it. This card requires no power connector and is tiny consuming only about 30W though.

MSI GTX 950 (2 GB RAM) (_Maxwell (2nd Gen) GM206 5th Gen NVENC chip_)

```auto
*------------------------------*---------------------*
| Concurrent Sessions | 6 |
| Total Segs Transcoded | 180 |
| Real-Time Segs Transcoded | 180 |
| * Real-Time Segs Ratio * | 1 |
| Total Source Duration | 360.00000000000017s |
| Total Transcoding Duration | 39.78670270800002s |
| * Real-Time Duration Ratio * | 0.1105 |
*------------------------------*---------------------*

```

Similar performance here, caps out at 6 concurrent streams thanks to its limiting 2 GB RAM.

Looking at the _Real-Time Duration Ratio_, it appears both nv{enc/dec} chips are capable of more concurrent streams, but run out of RAM:

```auto
[h264_cuvid @ 0x68432c0] ctx->cvdl->cuvidCreateDecoder(&ctx->cudecoder, &cuinfo) failed -> CUDA_ERROR_OUT_OF_MEMORY: out of memory
[h264_cuvid @ 0x68432c0] ctx->cvdl->cuvidDecodePicture(ctx->cudecoder, picparams) failed -> CUDA_ERROR_INVALID_HANDLE: invalid resource handle

```