Need help installing a GPU

105437

BIT Beta Team
Jun 8, 2015
2,159
1,051
I have a Dell OptiPlex 3010 that I have purchased an Nvidia Tesla K20 for. Is it pretty much plug-n-play or is their software/driver installs necessary? Thanks!
 
My GPU should arrive in the next day or so. Any advice or comments would be greatly appreciated. Thanks!
 
The GPU itself is just plug-n-play, but to do anything with it like offload DeepStack to it, there are some files you need to do as part of the DeepStack for GPU.
 
  • Like
Reactions: user8963 and 105437
First install these files and then download and install DeepStack for GPU:

Install DeepStack GPU
To install and use DeepStack GPU version on your Windows machine, follow the steps below
Once the above are installed, download and run DeepStack GPU version for windows via the link below.
>> Download GPU version for Windows


 
Excellent, I really appreciate it. Looking forward to see what kind of improvement I get. I'd really like to start using the dark model, but right now with no GPU and the CPU version, it takes 6 - 10 seconds to process images.
 
You may have to go to nviidia website and download the newest drivers for that card first. ? Usually the driver disc that comes with it is outdated already.
 
  • Like
Reactions: sebastiantombs
Well... I'm pretty bummed. I received the GPU today and apparently I need a low profile card to fit inside my Dell OptiPlex 3010 DT. The card fits just fine except it's too tall and I can't put the lid back on. Ugh!
 
You could always cut an opening out of the LID for the GPU card to stick out of :lmao:

There are not too many low profile cards that have a lot of horsepower. I use the low profile 1030 that gets its power from the motherboard and it isn't a powerful GPU by any means, but it was an improvement in DeepStack by 6-8 times.
 
  • Like
Reactions: 105437
Look at the Tesla series from NVidia. Low profile, single slot, card but installing a fan for it is a good idea.
 
  • Like
Reactions: 105437
CUDA cores are more important than memory from what I can see so far. 1,000 or more seems to provide the best bang for the buck and shortest analysis times.
 
  • Like
Reactions: 105437
That is the "equivalent" to the GT1030 I mentioned, except the 1030 has more CUDA cores.