SI
SI
discoversearch

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.
Technology Stocks : Advanced Micro Devices - Moderated (AMD) -- Ignore unavailable to you. Want to Upgrade?


To: mas_ who wrote (249071)3/20/2008 8:58:29 PM
From: b4dt4steRead Replies (1) | Respond to of 275872
 
45nm 6MB L3 Shanghai on HT3 links will be Nehalem's main competitor not Barcelona.

Both are future to-be-released products, but I believe the risk of a lithographic process change (in the former) is much greater than the risk of an architectural change (in the latter).

At the very physical (device) level, there are one, maybe two ways to solve a problem. Think about how few physical processes one can use to build a hard drive (using the GMR effect en.wikipedia.org ) or to build a transistor (FET or BJT). In fact, the solutions are so rare yet important that, for example, the inventors/discoverers of the GMR and the transistor were awarded Nobel Prizes.

On the other hand, a high-level architectural or algorithm problem has _many_ solutions. One example is encoding/decoding in communication protocols, where one can communicate by changing the frequency (FM, PSK, etc.), or by changing the amplitude (ASK, M-ary ASK, etc.) Sorting algorithms are another example -- there exist good sorting algorithms such as quick sort, heap sort, etc.

HT3 is a great link, but a company like Intel can come up with another one (CSI) that is just as great.



To: mas_ who wrote (249071)3/20/2008 9:19:40 PM
From: graphicsguruRead Replies (1) | Respond to of 275872
 
Is there confirmation that HT3 will finally
be a reality with Shanghai?



To: mas_ who wrote (249071)3/20/2008 9:27:40 PM
From: graphicsguruRead Replies (1) | Respond to of 275872
 

45nm 6MB L3 Shanghai on HT3 links will be Nehalem's main competitor not Barcelona.


How much improvement do you expect from Barcelona
to Shanghai?

How much improvement do you expect from Harpertown
to Nehalem?

Note that the examples you give where Barcelona is
competitive seem to be where either (a) Harpertown
is limited by fsb or (b) Harpertown is limited
by multithreading locks. Otherwise, Barcelona
gets trounced.

(a) gets fixed by the integrated memory controller in
Nehalem and 3 memory channels and QPI.

(b) gets fixed by the memory controller and QPI and
the fact that Nehalem becomes a "true quad core" and
by the fact that Intel has led us to expect much better lock
performance on Nehalem vs. Harpertown.

Where does that leave Shanghai's niche?