CUDA thread

Thread replies: 11
Thread images: 4

Anonymous
CUDA thread 2016-10-23 13:33:42 Post No. 57202970
[Report] Image search: [Google]

File: GeForce_GTX_1080_Front_86081672-334e-4f1e-a8c1-cffb3f451a61-prv.jpg (33KB, 700x367px) Image search: [Google]

CUDA thread Anonymous 2016-10-23 13:33:42 Post No. 57202970 [Report]

So which consumer-end card would be better for CUDA development on a budget?

I'm looking into:

GTX 1080:
$600
2560 CUDA Cores
8 GB GDDR5X
320GB/s

GTX 1070:
$400
1920 CUDA Cores
8 GB GDDR5
256GB/s

From a price/performance perspective the GTX 1070 seems like the better choice especially given that both have the same amount of memory, but I do want to hear opinions from other CUDA developers.

Anonymous 2016-10-23 13:45:46 Post No.57203073
[Report]

Anonymous 2016-10-23 13:45:46 Post No.57203073 [Report]

>>57202970
>>>/v/

Anonymous 2016-10-23 13:48:28 Post No.57203101
[Report]

Anonymous 2016-10-23 13:48:28 Post No.57203101 [Report]

>on a budget
1070

Anonymous 2016-10-23 13:52:02 Post No.57203144
[Report] Image search: [Google]

Anonymous 2016-10-23 13:52:02 Post No.57203144 [Report]

File: 1475854172510g.jpg (160KB, 1280x720px) Image search: [Google]

160KB, 1280x720px

Ditch CUDA for OpenCL.

Anonymous 2016-10-23 13:58:11 Post No.57203205
[Report]

Anonymous 2016-10-23 13:58:11 Post No.57203205 [Report]

This thread has my interest.
Until now I was using an Nvidia Jetson since I do embedded but I am thinking of /just putting a desktop GPU in the system/

Anonymous 2016-10-23 14:01:23 Post No.57203234
[Report] Image search: [Google]

Anonymous 2016-10-23 14:01:23 Post No.57203234 [Report]

File: 1476477982255.gif (2MB, 300x281px) Image search: [Google]

2MB, 300x281px

>>57202970
If you're only interested in CUDA, get the original Titan. It crushes the 1080 in double precision and was around $200 the last time I checked.

Anonymous 2016-10-23 14:03:07 Post No.57203251
[Report]

Anonymous 2016-10-23 14:03:07 Post No.57203251 [Report]

>>57203234

3-4 of these used

Anonymous 2016-10-23 14:12:11 Post No.57203341
[Report]

Anonymous 2016-10-23 14:12:11 Post No.57203341 [Report]

>>57203144
I'm far too comfortable with the various CUDA libraries out there at this point to really consider switching, but that is probably a pretty lame excuse...

>>57203205
I've done a bit of embedded myself, but only on platforms that don't have GPUs

>>57203234
https://www.amazon.com/gp/offer-listing/B00BJ7AJ7C/ref=dp_olp_used?ie=UTF8&condition=used

More than $200, but still looks like a solid option. Most of our work is done using single precision, but some of our potential ventures may require double precision. Thanks, I'll look into it.

Anonymous 2016-10-23 14:15:07 Post No.57203371
[Report] Image search: [Google]

Anonymous 2016-10-23 14:15:07 Post No.57203371 [Report]

File: 1476478554129.jpg (123KB, 972x687px) Image search: [Google]

123KB, 972x687px

>>57203341
I just checked and there are lots of them on ebay. It looks like $275 is around the standard price.

Anonymous 2016-10-23 14:17:11 Post No.57203389
[Report]

Anonymous 2016-10-23 14:17:11 Post No.57203389 [Report]

>>57203371
Ah, thanks, I should have checked that site out too

Anonymous 2016-10-23 15:16:31 Post No.57203988
[Report]

Anonymous 2016-10-23 15:16:31 Post No.57203988 [Report]

GTX 1050 Ti if you don't want to spend much and just want to dip in toes first into CUDA

I'm aware that Imgur.com will stop allowing adult images since 15th of May. I'm taking actions to backup as much data as possible. Read more on this topic here - https://archived.moe/talk/thread/1694/

I'm aware that Imgur.com will stop allowing adult images since 15th of May. I'm taking actions to backup as much data as possible.
Read more on this topic here - https://archived.moe/talk/thread/1694/