Date: Thu, 08 Aug 2002 11:55:15 +1000 From: Rob B <rbyrnes@ozemail.com.au> To: freebsd-alpha@freebsd.org Subject: Re: not dead yet? Message-ID: <5.1.0.14.2.20020808113351.01dc3100@pop.ozemail.com.au> In-Reply-To: <15697.22873.215915.596918@grasshopper.cs.duke.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
At 03:31 8/08/2002, Andrew Gallatin sent this up the stick: >http://www.theinquirer.net/?article=4842 There's an interesting thread on this at the moment on RedHat's axp-list mailing list: Quote ="David Mihail" >As for Samsung's various Alpha models, the one with 1.75MB on-die L2 is the >21264E, it's done in .18um, not .13um, which makes it an EV68 variant, not >EV69. Also, it already runs at 1.25GHz. And that's using Aluminum instead >of Copper, and no SOI. It's also a big chip, ~300mm2, so it's probably not >going to be cheap, especially considering that Samsung gets lousy yields on >"standard" Alphas. And I mean lousy. And going from a 115mm2 die to a >300mm2 die is just going to make their yields worse, which will jack up the >price even more. You can't really blame them, though, Samsung specializes >in manufacturing SRAM/DRAM/Flash and LCDs, not microprocessors. It's a >little bit different. > >As far as the EV69 goes, it's actually an IBM part. It's done in .13um with >both Copper and SOI. Takes up a die space of ~60-70mm2. It runs at >~1.5GHz. Considering that nobody, not even Intel, makes processors as well >as IBM does, and taking the smaller die into consideration, I'd say that >yields aren't going to be as much of an issue. Can we get our hands on >them? I don't know. Might not do us much good if we could, it might not be >a drop in replacement. *Standard* EV6x is in a 587-pin package. EV69, I've >heard (which definitely makes it a rumor) is in a 600+ pin package. We also >don't know whether IBM is allowed to sell them to anyone other than HPQ. > >*Can* Samsung continue producing Alpha for as long as they want and make >whatever changes they want to it? Yes they can. *Will* they do so? Well, >that's what I'm working on. Everybody here needs to understand that Compaq >bent them over a barrel and didn't even have the common courtesy to give >them a reach around. Sorry for the vulgarity, but that's the situation. >Have you ever seen a dog that's been repeatedly and sorely beaten? Do you >think that dog is going to trust you and come when it's called? Even if >you're the new owner and have never touched it, do you think it'll trust you >now? No, it won't. It takes a while to build up that trust again. > >So what's the point of that? Samsung needs it to be worth their while to >continue fabbing Alpha. Since my name's already been mentioned, sure, I'll >confirm that I'm doing absolutely everything that I can to see that we get >new Alphas. But it takes time. And we have to take care of Samsung. The >problem is that there's a certain volume that they need, and the simple fact >of the matter is that under the current model of using Alphas in mid- to >high-end servers, with no new Alphas on the roadmap, it's just not going to >happen. We're talking a full order of magnitude more of sales. There's >only one way that it will happen, and that's to make Alpha the heart of a >general purpose computer (that's right, I just said it needs to be a PC, an >AlphaPC, like the MacPC, but not sucky), and you need something other than >Linux for that. You can disagree with me all you want, but it's fact. >Witness that Alpha is end of life. Cheers, Rob PS: Apologies - sigmonster has been bad today :) -- We should build an Intel processor out of penguins. This is random quote 1147 of a collection of 1254 [15200.8 km (8207.8 mi), 262.8 deg](Apparent) Rennerian To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-alpha" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?5.1.0.14.2.20020808113351.01dc3100>