Comments on: Scott Clark on EDA Clouds http://theasicguy.com/2010/08/08/scott-clark-on-eda-clouds/ sharing insights into the people side of ASIC design Thu, 17 May 2012 16:28:58 +0000 http://wordpress.org/?v=2.3.2 By: Scott Clark http://theasicguy.com/2010/08/08/scott-clark-on-eda-clouds/#comment-2147 Scott Clark Wed, 18 Aug 2010 04:16:49 +0000 http://theasicguy.com/2010/08/08/scott-clark-on-eda-clouds/#comment-2147 James, I agree that the bar will continue to move on refresh performance, and the question we always face is not what is "fastest" but more of what is "acceptable" from a user experience perspective (and that is qualitative and specific to the user). One thing that is coming though is much higher bandwidths at far more reasonable prices (still does not change latency, but if congestion is an issue, will definitely change the experience). The challenge with this (getting much more for the same price) will be to get IT and/or the CFO to accept the performance increase for the same price instead of trying to get credit for a cost "savings" while maintaining the same performance :) I would be really interested in hearing more about your display benchmarks, that sounds like good stuff... -swc James,
I agree that the bar will continue to move on refresh performance, and the question we always face is not what is “fastest” but more of what is “acceptable” from a user experience perspective (and that is qualitative and specific to the user). One thing that is coming though is much higher bandwidths at far more reasonable prices (still does not change latency, but if congestion is an issue, will definitely change the experience). The challenge with this (getting much more for the same price) will be to get IT and/or the CFO to accept the performance increase for the same price instead of trying to get credit for a cost “savings” while maintaining the same performance :)

I would be really interested in hearing more about your display benchmarks, that sounds like good stuff…

-swc

]]> By: James Colgan http://theasicguy.com/2010/08/08/scott-clark-on-eda-clouds/#comment-2112 James Colgan Mon, 09 Aug 2010 19:32:05 +0000 http://theasicguy.com/2010/08/08/scott-clark-on-eda-clouds/#comment-2112 Nicely done Scott. Agreed on all points. I particularly like your point that the industry should look to the foundry model for inspiration (<a href="http://www.xuropa.com/blog/2009/03/26/the-cad-less-semiconductor-company/" rel="nofollow">The CAD-less Semiconductor Company</a>). We've been down this path before. Regarding performance - fortunately, there are many companies and industries out there that their business depends on continually raising the bar on this front. Screen refresh is cited as an initial concern by our customers, but side-by-side benchmarking done by <a href="http://www.xuropa.com/cadence" rel="nofollow">Cadence</a> on Xuropa have shown comparisons of between 98% and >100% of local performance. Public Cloud vendors will likely lag what semi IT departments have built in terms of cluster performance for a while, but that's fine. Not everyone needs that level of performance all of the time. (One size does not need to fit all.) The longest pole in the technical user experience tent will continually be transfer rates. High-Definition Video-on-Demand proliferation will help stimulate solutions to this problem, but it will be a while before we see 1TB "zipping" over the wire ;-) Nicely done Scott. Agreed on all points. I particularly like your point that the industry should look to the foundry model for inspiration (The CAD-less Semiconductor Company). We’ve been down this path before. Regarding performance - fortunately, there are many companies and industries out there that their business depends on continually raising the bar on this front. Screen refresh is cited as an initial concern by our customers, but side-by-side benchmarking done by Cadence on Xuropa have shown comparisons of between 98% and >100% of local performance. Public Cloud vendors will likely lag what semi IT departments have built in terms of cluster performance for a while, but that’s fine. Not everyone needs that level of performance all of the time. (One size does not need to fit all.) The longest pole in the technical user experience tent will continually be transfer rates. High-Definition Video-on-Demand proliferation will help stimulate solutions to this problem, but it will be a while before we see 1TB “zipping” over the wire ;-)

]]>