Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Politics : Formerly About Advanced Micro Devices -- Ignore unavailable to you. Want to Upgrade?

To: AK2004 who wrote (30981)	4/4/1998 1:25:00 AM
From: Kashish King	Read Replies (3) \| Respond to of 1573514

This guy rode SIII from $25 to $5 and RACE from $25 to $1.75 but do you think he might be right about this: Apr 4 1998 1:16am I wouldn't say gone - AMD has made a fine CPU for years yet few will buy anything that can't claim a Pentium processor. Marshall He's big on marketing brochures. After almost two years, they haven't sold anything but some noise level pilot installations of a product that hooks up old PBX systems to Lucent modems with some duct tape, bailing wire and a little software: exchange2000.com

To: AK2004 who wrote (30981)	4/4/1998 1:48:00 AM
From: Ali Chen	Read Replies (1) \| Respond to of 1573514

Albert, <Elbrus will ... be more effective in computational operation than Merced.> I would not worry about Merced at all. The Merced is (?) a product of design behind tightly closed doors, by a group arrogant and overpaid internal engineers. High secrecy and inability to discuss ideas inside wide scientific community cannot give a fruitful result. Just looking at the design idea - three explicitly parallel instruction to execute in three units - so what? Looks impressive at first glance - 3X instant leap in performance, but in reality? As for any serious scientific computation this parallelism by three is nothing, nil. For server-like applications, a single processor is more than enough to route the communication packets. You may think that all three units could do three times more work. Wrong. These requests are most likely uncorrelated in time, and you would need a sort of DYNAMIC compiling all the time to combine instructions in triades to feed this under-brained flop. It will lead to constant cache trashing, with huge associated memory traffic, and zero effective performance. In addition, three processors is apparently not enough to handle 40-50 service processes in NT, so you still need a fast time sharing capability. Flop here again: you cannot save the enormous amount of registers and switch the task state fast, just because it would be too much to save and restore. Remember, Intel was always the worst in task switching and real-time response, how they could design anything useful without experience? I believe they even are unaware of this problem. As far as desktop/workstation concern, you know better: "all PCs wait at the same rate". So, take care. Ali, always with a screwdriver, and bubbling about:)