Will Cedar Mill save face for Intel?

knewt

Distinguished
Oct 6, 2002
132
0
18,680
I finally found some worthwhile info on Intel's next desktop CPU in the works. Although some of the info is quite out of date now, it has some very interesting details.

We know the processor will actually be dual core. I wonder how this will affect the power output. 90W sounds so much better than what we have on Prescott.

Intel is approximately 6 months ahead of schedule in 65nm development. I expect we will see this chip launched ahead of schedule. Granted this would be certainly be a refreshing change from recent events. I am predicting we will see the first shipments of Cedar Mill by September 2005.

This chip should be quite an improvement. I do know it will include all of what Tejas was going to have plus more. Hopefully the performance will be really competitive again.

<A HREF="http://www.x86-secret.com/index.php?option=newsd&nid=762" target="_new">Clicky</A>

<font color=white>-</font color=white>
Wanted: Large breasted live-in housekeeper. Must be a good cook, organized, and willing to pick up after me.
 

endyen

Splendid
Well, that 90 watts is what they are hoping for with prescott on 65 nanos,with a single core. Doesn't seem that great to me. They say it will be dual core ready, but how many watts would that be? There is also mention of a dual core xeon, but the first dual core shown on the roadmap is Jonah2, in 06.
By the way, Intel is ramping up for 65 nano because they are trying to find a way of avoiding licensing IBM's soi tech. They are also expecting severe problems, and are looking for time to fix them.
 

knewt

Distinguished
Oct 6, 2002
132
0
18,680
Well, that 90 watts is what they are hoping for with prescott on 65 nanos,with a single core.
Check again, 90 watts is for Cedar Mill. Prescott lists TDP at 113W.
They say it will be dual core ready, but how many watts would that be?
My point exactly, don't know what dual core will do to the power issue, but the original single core was obviously being designed to be much cooler than Prescott.
There is also mention of a dual core xeon, but the first dual core shown on the roadmap is Jonah2, in 06.
As I stated, this info is quite out of date. I don't remember what the updated forcast for dual core xeon is supposed to be, but I imagine it will be close to the same time as Pentium.

Upon closer inspection, I noticed Yonah's power consumption is 23 and 31 watts. This is not likely the Xeon chip. I forget all the codenames, but based on the power, I'm thinking it must be the Pentium M. Correct me if I am wrong.
By the way, Intel is ramping up for 65 nano because they are trying to find a way of avoiding licensing IBM's soi tech. They are also expecting severe problems, and are looking for time to fix them.
Intel's 65nm process development is so far ahead of schedule because it is going much smoother than anticipated. I don't believe that has anything to do with IBM's SOI tech. They have already decided not to go with SOI.

Every new generation is expected to have severe problems to work out but things are going better than expected this time. The point is they are likely going to fix the problems in less time, not more.

I just hope their next generation of processors are going to be more competitive.

<font color=white>-</font color=white>
Wanted: Large breasted live-in housekeeper. Must be a good cook, organized, and willing to pick up after me.<P ID="edit"><FONT SIZE=-1><EM>Edited by knewt on 08/13/04 10:31 PM.</EM></FONT></P>
 

endyen

Splendid
It would be nice if your rosie colored glasses came true.
Cedar mill is the new name for prescott on 65 nanos.
Jonah (Yonah in hebrew, or french) is Dothan on 65 nanos.
Rumour has it that TDP for cedar mill duals would be 150 watts, but it may be closer to 180, depending on leakage.
Intel has begun ramping early for 65 nanos, thier spinmasters translate that to being ahead of schedual. The fact is, Intel is running scared. The extra 6 months should help get them back on track. I hope so, but exacution hasn't been great lately. I was realy hoping for a stepping that would drasticly reduce the heat from scotty. It's not that the current numbers are that bad, it's just not leaving a lot of room for ramping. Changing to 90 nanos doesn't sound all that great either.65watts puts them back to P4c 3.2 TDP.
Oh well I guess I'm a pessamist.
 

Mephistopheles

Distinguished
Feb 10, 2003
2,444
0
19,780
I think an early transition to 65nm, still in 2005, would probably help intel immensely. The problem with current Intel tech is not that bad really, but prescott more than doubled the amount of logical transistors in prescott, which was horrible for thermal output (heck, 2 NW cores would fit in there...)

I wouldn't put too much confidence in that roadmap... Intel's probably shuffling desperately through their cards in search of a trump card. They need it fast. Until they find it, I'd expect their roadmap to be vague at best.

And Endyen, it would be nice to see scotty dissipating, say, 20-30W less, but I'm thinking that that would only maskerade the problem. Intel should know better and switch to a saner design. A performance-oriented one, through balanced increase of IPC and Mhz, instead of an unbalanced mix of a lot of clock and very precious little IPC.

Forget Netburst, Intel, it's a bust. As far as we all can see it.
 

knewt

Distinguished
Oct 6, 2002
132
0
18,680
One thing you can be assured of is that Intel knows it has a problem with heat and they are working very hard to find a solution. I hope you are wrong about 150w for dual core. That wont be a step in the right direction.

I do know they are working on a completely new architecture. I've heard rumers that it is targeted at around 45w, but don't expect to see it until 2006. Until then we can expect more varieties of Netburst.

<font color=white>-</font color=white>
Wanted: Large breasted live-in housekeeper. Must be a good cook, organized, and willing to pick up after me.
 

Mephistopheles

Distinguished
Feb 10, 2003
2,444
0
19,780
One thing you can be assured of is that Intel knows it has a problem with heat and they are working very hard to find a solution.
Well, at least in theory, Intel should have a significant number of think tanks available and working on it. As long as their marketing department doesn't work against those think tanks (e.g. "give us more Mhz"), then Intel should manage to put their act together at some point.
I do know they are working on a completely new architecture. I've heard rumers that it is targeted at around 45w, but don't expect to see it until 2006. Until then we can expect more varieties of Netburst.
Were those net based rumors? I have not read anything about a 45w architecture or so. Too high for dothan, too low for netburst. Do you have additional information?
 

knewt

Distinguished
Oct 6, 2002
132
0
18,680
Yeah, the code name for it is Merom. Don't know how accurate the info is, but it looks rather legit.

<A HREF="http://endian.net/roadmap.asp?IxStart=25" target="_new">Clicky</A>

<font color=white>-</font color=white>
Wanted: Large breasted live-in housekeeper. Must be a good cook, organized, and willing to pick up after me.

My bad, that is actually a mobile chip, Conroe is based on it and will be for desktop at around 90W again.<P ID="edit"><FONT SIZE=-1><EM>Edited by knewt on 08/14/04 08:25 AM.</EM></FONT></P>
 

juin

Distinguished
May 19, 2001
3,323
0
20,780
65NM equipement as allready been buy.A 30 to 20 watt reduction is about the same that 130 nm have bring over 180 nm.
The thing is 90 nm work dothan have decrease is tpd and încrease is cache size with higher clock.Intel have face leakage in the dielectric not a ground effect.High k eff is what is need.For the growing ground effect Intel can allwayse come with Sillicon on air is work well and increase cost by a very small margin as air cost nothing.Toshiba have look on this avenue.There a issue with any SOI as it very hard to come with a strain from the isulator there 2 tech on this SI-GE on isulator or SI on Isulator the last option is the best but no one have able to make on mass production unless there something i dont know.

By the way IBM dont have the monopole on SOI as they buy the wafer like everyone.There a association of PD-SOI and FD-SOI wafer manufacturer they hold most patent.IBM hold some manufacturing patent on SOI wafer.Also intel was the 1 to have try PD-SOI and FD-SOI in lab.

i need to change useur name.
 

endyen

Splendid
I think you hit the nail on the head when you said the extra transistors wre the problem.
I think that Northwood with 1 meg of cache, EM64T registers, a 266/1066 fsb on 90 nanos, would be a much better solution than scotty ever was. It would easily scale past 4 gigs, and with 65 nanos, it should pass 5 gigs. If the fsb can be brought to 300/1200 all the better.