Winchester Rumors

[amol]

I'm hearing rumors on the internet that the new 90nm Winchester will either be short lived or upgraded soon.

There is speculation that SSE3 might be added to another 90nm core.

Anyone hear about this, and if its true, around when these new CPUs would come out?

Also, is SSE3 really important?

 

[ChunkyBarf]

Amol: Concerning your question about SSE3 being important: The more important question is when. Sure, eventually, extra SSE3 instructions will be important, but not until software is optimized for it.

I cannot comment on the other aspect of your question (90nm revisions). I do not know anything about the situation. I cannot conceive of a core revision yielding tremendously different performance though if history is any indicator.

Perhaps someone more knowledgeable in the subject can elaborate further.

Good luck,
ChunkyBarf

 

[jkresh]

SSE3 is coming, but not until Q1 2005 (last I saw anyway), sse3 makes a big difference with some apps (media encoding mostly) and small difference to no difference with many. If you want to upgrade now then buy something cheap 3000+ or 3200+ and if sse makes a big difference for you upgrade and sell the old chip in Jan/Feb/Mar whenever it comes out.

 

[uOpt]

Actually SSE3 doesn't have much for media encoding, most of the new instructions are helpers for geometric programming.

 

[UUronl]

"The upcoming AMD San Diego and Venice core AMD64 processors - Athlon 64 FX and Athlon 64 respectively - will feature support for Intel's SSE3 extended intruction set. Even more important is planned optimizations for the memory controller to improve the performance of motherboard chipsets with integrated graphics.

The first Revision E CPUs will begin shipping in early 2005. AMD also has plans to introduce an Athlon 64 4200+ towards the middle of 2005; they are not listing whether the part will feature a 512KB or 1MB L2 cache, but it will most likely run at 2.6GHz. The Athlon 64 FX-57 is also listed on the roadmap as a 2H-05 part, it's specifications are also unclear but we'd expect it to be a 2.8GHz part with a 1MB L2 cache. Both the Athlon 64 FX-57 and Athlon 64 4200+ appear to be 90nm parts built on the San Diego and Venice cores, respectively."

 

[amol]

actually, i was looking up amd's CPU roadmap, and the toldeo core interests me . . .

anyone know when it is coming out? if it is more than a year (and if it is S939), then i might as well get a winchester right now, then upgrade when toledo comes out . . .

 

[Thermalrock]

you should better hurry and get a winchester before youre stuck witha venice. winchester sounds cooler than venice, enough to easily make up for the sse3 lack. i think sse3 isnt that important, just the prescotts and noconas have it. northwood pentium m and even the p4 extreme edition, that intel offers for 1000+ bucks dont have it. also i read somewhere about sse2 that a64 and pentium m dont benefit from it as much as p4 because of the clockspeed. now if this is true i think it would apply to sse3 too.

 

[Seemann]

AFAIK the Winchester will be equipped with following goodies:
* SSE3
* "Adaptive Prefetch" to optimzie cache-hit-rate
* two additional Write-Combining Buffers (then 4 WBCs)
* XOR DRAM Bank Address to optimize cache write-back strategy
* extended"Clock Ramp Hysteresis Counter"

These features are planned for the new E-stepping of A64-core, the production will probably start in the beginning of 2005. So we can assume that these chips are buyable somewhere in the middle of 2005.

Here's the source:
http://www.planet3dnow.de/cgi-...=all&id=1099563303

 

[clarkey01]

In the first half of next year, AMD will shift its FX and A64 chips to the San Diego, Venice and Palermo 90 nanometre SOI chips, displacing existing .13µ (micron) processors


Inq..

 

[itachi]

Originally posted by: MartinCracauer
Actually SSE3 doesn't have much for media encoding, most of the new instructions are helpers for geometric programming.

sse3 has a lot for media encoding. one of the new instructions, lddqu, does wonders with reducing the computation time of motion estimation, which is the main contributing factor to the encoding time for video.
then there's the additional instructions for complex arithmetic. from intel's site, "Complex arithmetic usage is ubiquitous, as it is used in Discrete/Fast Fourier Transform (DFT/FFT), Discrete Multi Tone (DMT) modulators, frequency domain filtering, etc." seems like sound-related applications will be the one that benefits the most from this. not only sound tho.. any app that performs calculations in the frequency domain should benefit from this.
mp3 encoding, once implemented, will make a lot of use of the new instructions, not that its ever really been an issue.. it benefits from the same instruction as video does and for the same reason and the bulk of the encoding process involves waveform analysis (perceptual coding and bit allocation).

just out of curiosity.. but, what instructions were you referring to when you stated that most of the new ones were helpers for geometric programming? no clue what geometric programming is except that it's used for cad. the horizontal instructions were made for graphics programming.. no clue about complex arithmetic, but i refuse to believe that the reason for its existance is in any way related to geometry.