Prescott to have Dual channel DDR 333 controller On board!

[Pentium]

http://www17.tomshardware.com/mainboard/02q2/020624/i850e-24.html hehe!! Intel just Killed Hammer's Main advantage!

 

[Rainsford]

Fanboys Unite!

I plan on waiting till the new CPU's are out to make any decision about which is the best. Specs really don't mean a whole lot as a lot of us have learned the hard way. Many products look great until they actually come out. Not saying this will happen to Prescott or Hammer, but we know very little about how they will perform in real world applications.

 

[Vic]

Will the Prescott support 64-bit? No? Then it hasn't killed the Hammer's "main advantage."

 

[imgod2u]

I would say Hammer's main advantage is the integrated memory controller plus some of the scheduling optimizations made on the core itself. The x86-64 support is just a bit short of hype. There isn't really any consumer-level software that would benefit at all from it. And that's what Prescott is to compete with Hammer in, the consumer market. With Madison and Deerfield, Intel's bring IA-64 to the low-end server and workstation markets, so there'll be no more need for Xeon.

 

[Vic]

People said the same thing when the 386 came out. "Why do I need 32-bit, there are no programs available for it, no one will ever need that much RAM... "

 

[MajinVegeta]

On board not on die. hammer will have a lanecy advantage. for X86-64 to work AMD has to actaully support compliers for it not hope Intel does for them like with MMX and SSE.

Plus x86-64 software will be a rarity in the destop market for the first 2 years of launch in the 4-8 way MP it will not but deskto users wont see any edge becuase there isnt any reason to swith 64 bit pointers take 8x the bandwith over 32 bit pointers.

 

[Pentium]

On board, I meant "On-Die" Satisfied? anyway The Prescott has the Instructions on them but not enabled. So Intel will wait and see if 64 bits are needed, If so You can bet Intel will put them on!

 

[Accord99]

Originally posted by: PSYWVic
People said the same thing when the 386 came out. "Why do I need 32-bit, there are no programs available for it, no one will ever need that much RAM... "

And how long did it take before 32-bit software became mainstream? Nearly a decade, by which time the 386 had long been obsoleted.

 

[bunnyfubbles]

and how drastically did hardware advance in that decade? Hmm...

 

[imgod2u]

Well, as always, the transition to 64-bit will be needed eventually. However, not for 2 years I'd say. Currently, there is no consumer-level software that would benefit from 64-bit extensions. And no boards will even support 2 GB of memory, let alone 4 GB and nobody's complaining. At least, not in the desktop market.
This is why Hammer's main advantage will be in the workstation market.

 

[MajinVegeta]

I dont get why everyone here is saying 64 bit for memory addressing, becuase hammer uses Xeons 48/40 bit addressing, not a true 64 bit memory addressing system. The xeon 48/40 bit addressing setup supports 280 TB of ram where a real 64 bit soultion would support 18,000 TB.

 

[imgod2u]

The difference is, Hammer will have flat 48-bit memory addressing while the Xeons require an extended window. Flat memory addressing is much better.

 

[mechBgon]

I wonder if we will see fewer stones being thrown at AMD for their on-die controller, now that Intel is known to be taking a similar approach. edit: ah, probably not. I see that by "on board" we are talking motherboard-based... never mind.

 

[CrazySaint]

linkified

 

[yokem55]

Or Intel could put prescott and a dual ddr333 northbridge onto a slot type package and the memory controller would then be "on board". Doing this could remove a lot of the latency from having an external memory controller as Intel would be able to control the design and layout of the fsb completely. It would also make their chipset people very happy.

 

[Rainsford]

Originally posted by: yokem55
Or Intel could put prescott and a dual ddr333 northbridge onto a slot type package and the memory controller would then be "on board". Doing this could remove a lot of the latency from having an external memory controller as Intel would be able to control the design and layout of the fsb completely. It would also make their chipset people very happy.

I would think this might create somewhat of a marketing backlash. I assume there was a good reason to go with socket for modern CPUs over slot, but even if there wasn't, going back to "old" technology wouldn't look too good I would think.

 

[imgod2u]

The 486's and Pentiums were of socket design. Then the advent of L2 cache. When put closer to the processor, it improved performance. So the slot design was thought up. Then, manufacturing processes got to the point where the L2 cache could be put on die, so the socket design was brought back. Now the trend is to move the memory controller closer to the processor so we may very well move to a slot-type design again. And then later, as manufacturing processes again improve, move the memory controller on-die and go back to socket designs.

 

[Pentium]

Slot? Not gonna happen. Prescott is to use socket 478 and If AMD can put it on a socket, Intel with their superior manufacturing process will be able to put it on board easily.

 

[CrazySaint]

I would think that the biggest issue with a slot-type design now days would be cooling. There could be issues with large HS/F and airflow.