TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Rambus, Microsoft Put DRAM into Deep Freeze to Boost Performance

34 点作者 RmDen大约 8 年前

6 条评论

ChuckMcM大约 8 年前
Heh, I thought &quot;wait a minute didn&#x27;t IBM do this 20 years ago and it was a total flop?&quot; and yes, and its one of the same people involved.<p>Cryogenic computing wasn&#x27;t a bust because the technology didn&#x27;t work. It does. It was a bust because silicon <i>really</i> hates to transition between cryogenic temperatures and room temperature. If you transition it quickly (say you pull a card out of the liquid nitrogen and start working on it) it will crack as it warms up unevenly. As a result you needed anywhere from 20 to 48 hours to get a card from &#x27;cryo&#x27; temp to &#x27;room&#x27; temp, and while you could cool a bit faster it still took longer than just dumping it in LN2. So repairs and maintenance were multi-day affairs. Compare that to a modern AWS, Google, or Azure data center where a system fails, a tech can skate out to it with a new motherboard, pull the old one, put the new one in, and poof you&#x27;re back on line in under 30 minutes.<p>As a result cryo computers either had to have failure rates that were so low that a repair that required transitioning the hardware through a cold&#x2F;warm&#x2F;cold cycle rarely happened, or you had to have enough extra hardware to support your base load while part of it was slowly going through the cold&#x2F;warm&#x2F;cold cycle.<p>I don&#x27;t know how much IBM spent, but it was a <i>lot</i> and they never cracked that nut.
评论 #14288530 未加载
评论 #14290750 未加载
评论 #14288353 未加载
JohnBooty大约 8 年前
&gt; “Something not many people are talking about is playing the temperature card,” Bronner says [...] &quot;All the talk I hear is about stretching Moore’s Law in the datacenter, such as using GPU accelerators. Those will work, but it’s a one-time shot. You use it once and you are done.”<p>But isn&#x27;t &quot;the temperature card&quot; something you only get to play once as well? Once you&#x27;ve adopted supercooled computing, where do you go from there?<p>(This guy knows there&#x27;s such a thing as absolute zero, and that you can&#x27;t just keep getting colder, right?)<p>Sorry, I guess I still have antipathy towards Rambus since the 1990s. =)
评论 #14286565 未加载
评论 #14287490 未加载
frik大约 8 年前
Is Rambus still a thing? I remember their vendor lockin RAM memory from Pentium 4 era. Later there were some bad PR and lawsuits, and the PC world moved on. <a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Rambus" rel="nofollow">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Rambus</a>
评论 #14286395 未加载
评论 #14286381 未加载
评论 #14286851 未加载
评论 #14285952 未加载
评论 #14287601 未加载
davidgerard大约 8 年前
Is this the same Rambus that essayed into patent trolling standards? If so, I&#x27;m surprised they didn&#x27;t change their name.
评论 #14285862 未加载
petra大约 8 年前
Why would cooling offer such big performance improvement to DRAM ? even if it improves transistor witching speed, wire switching speed shouldn&#x27;t be affected and it is a significant component of latency , right ?
评论 #14286278 未加载
评论 #14285874 未加载
jdonaldson大约 8 年前
If they were smart, they&#x27;d advertise this with : &quot;Tech Ops HATES him! Check out this one weird trick that increases your memory speed 200%!!!&quot;