//flex table opened by JP

Click to See Complete Forum and Search --> : Unreleased core photo of a .13 P3 Tualatin


daveleau
06-20-2001, 12:23 PM
VERY cool stuff. Sysopt does it again. Thanks Scott/ RobRich!

Undeadlord
06-20-2001, 05:34 PM
I guess I am just stupid : but that picture makes no sense to me. Does the average or even above average user, really have any idea what that picture means?


Undeadlord

gr8vfr
06-20-2001, 06:15 PM
I have to agree...
The colors are very pretty, though.
But, could you explain what it all means?

Thanks.

randy48
06-20-2001, 06:32 PM
You don't know what the different colors mean? Come on, look close. Think of it as looking down at a small city, do you see the streets? The blocks are blue...that's were the Intel "Blue Dudes" grew up http://www.sysopt.com/news/gws/cdata/wink.gif

OK, so I'm stupid too! But I couldn't resist!

RobRich
06-20-2001, 07:19 PM
I didn't major in EE, but I'll try to describe a few of the basics areas. The two large light blue blocks in the left regions are L2 cache memory blocks, each consisting of 256kb of storage. The smaller light blue blocks in the same area are the L1 cache regions. Most other uniform blocks with similar consistancy should be related to register regions, buffering, and branch prediction tables.

The various transistors paths and related parts compose the instruction pipeline, ranging from decode to execute to memory i/o. If I could offer any better exaplanation of these regions' physical properties, then I would be at work for Intel. http://www.sysopt.com/news/gws/cdata/wink.gif

Robert Richmond

RobRich
06-21-2001, 01:50 AM
Thanks for the positive comments Dave. http://www.sysopt.com/news/gws/cdata/smile.gif

Upon examination of the core photo, notice the large cache storage areas. Also notice the close proximity of the controller stages to the cache region due to the smaller .13-micron die size. It appears the pre-release benchmark numbers floating around the net are correct, as the smaller length trace routes as compared to the .18u Coppermine could lead to a small reduction in latency. Overall, it appears to be a rather impressive design with great detail on efficiency and production yeild. I personally can't wait to see how the final revision Tualatin 512 performs against the Pentium 4. http://www.sysopt.com/news/gws/cdata/smile.gif

Catch ya' later,
Robert Richmond

zferenczy
06-21-2001, 05:49 AM
Thanks for the explanation Robert. We all think we know it all, but when it comes down to it we know a very little.

beard
06-21-2001, 07:02 AM
ok, where can i see an identical image but of a PIII for comparison? http://www.sysopt.com/news/gws/cdata/smile.gif

MTAtech
06-21-2001, 07:57 AM
It reminds me of when President Kennedy was given the aerial photos of Cuba, proving that missles where being installed. He relied entirely on the expert's interpretation of the photos since he couldn't make head nor tail out of it himself.

My hat is off to those that can make sense of this photo.

daveleau
06-21-2001, 09:27 AM
We all think we know it all, but when it comes down to it we know a very little.

Hey, speak for yourself, man. I DO know it all. http://www.sysopt.com/news/gws/cdata/wink.gif

Sorry, couldn't resist.

What about registers. Are those synonymous with cache or are registers elsewhere?

Dave

RobRich
06-21-2001, 12:53 PM
For comparison, P3 Coppermine core photos are available through the media relations portion of Intel's website.

http://www.intel.com

Registers are basically high-speed memory segments used for instruction storage within the execution pipeline. Check this out:

http://www.sysopt.com/articles/usparc-3/index4.html

Hope this helps,
Robert Richmond

zenjive
06-22-2001, 11:08 AM
Does anyone know where I could find more core photos like this? Maybe some AMD cores or something in a nice shade of RISC? MIPS, Sparc... Cray?

On the other hand, it will probably all look the same, hehe.

chass
06-23-2001, 12:13 PM
Are those features in the left half of the chip which appear to be solar panels, the 512KB of cache.I may be wrong but if the turns out to be the case, is half the chip is devoted to cache? anyone know how to decipher this?

krazyknuc
06-25-2001, 08:42 AM
Oh, you're all ****-dumb !
Can't you see the little whatchamacallits that do the processing and the thingamabobs that are cache and the doohickies that, OH, what's the use, you'll never understand this technical jargon ! http://www.sysopt.com/news/gws/cdata/smile.gif

RobRich
06-25-2001, 12:51 PM
All your questions are answered here (http://www.sysopt.com/forum/Forum17/HTML/003705.html) .

Hopen this helps,
Robert Richmond