<br><div class="gmail_quote">On Wed, Mar 24, 2010 at 6:50 PM, Paul Gray <span dir="ltr"><<a href="mailto:gray@cs.uni.edu">gray@cs.uni.edu</a>></span> wrote:<br><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
<div class="im">Anoop Saldanha wrote:<br>
<br>
<blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
i think gpus come with 2GB now. 1GB should be good enough though. It would be nice to go for a dual GPU graphics card, since we'll be looking at supporting load balancing on multi-cuda-devices at some point in the future. I think something like a gtx 295. You can look at a couple of other such versions in the market.<br>
<br>
</blockquote>
<br></div>
If there's an interest in Nvidia Tesla C1060's GPGPU hardware, we have systems with three C1060s (720 cores total, 4GB per GPGPU for 12GB on-GPU memory) that I can offer up for a test environment.<br>
<br>
We also have units with 4 GPGPUs total that could be used for testing, but not of the caliber of the C1060s.<br><font color="#888888">
</font><br></blockquote></div><br>I think that would be really useful, esp when we start to benchmark/test with different cards and multiGPUs. Thanks Paul<br clear="all"><br>-- <br>Regards,<br>Anoop Saldanha<br><br>