Northwood on a 0.9 micron process ?

[clarkey01]

With all the problems with preshott, would it been a better idea to have shurnk the northwood down onto a 0.9 micron process and slapped an extra 512 cahce on it ? I still think the core had more head room, just not on 130 nm.

 

[AristoV300]

The northwood's architecture maxes out at 3.6 regardless of the size of the core.

 

[clarkey01]

ah

 

[Alkaline5]

Originally posted by: AristoV300
The northwood's architecture maxes out at 3.6 regardless of the size of the core.

I'm no CPU engineer, but doesn't a die shrink almost always bring increased clock speeds? What about AXP .18 micron -> .13? T-Bred and Barton were nothing more than die shrinks (with very minor CPU alterations compared to Prescott) and they've now added about 1 GHz over the Palomino core.

Based on that assumption and the difficulty Prescott has given Intel, I think that a 90n Northwood core couldn't possibly have done worse for them.

 

[MrK6]

Originally posted by: Alkaline5

Originally posted by: AristoV300
The northwood's architecture maxes out at 3.6 regardless of the size of the core.

I'm no CPU engineer, but doesn't a die shrink almost always bring increased clock speeds? What about AXP .18 micron -> .13? T-Bred and Barton were nothing more than die shrinks (with very minor CPU alterations compared to Prescott) and they've now added about 1 GHz over the Palomino core.

Based on that assumption and the difficulty Prescott has given Intel, I think that a 90n Northwood core couldn't possibly have done worse for them.

Agreed :thumbsup:

 

[iwantanewcomputer]

Originally posted by: MrK6

Originally posted by: Alkaline5

Originally posted by: AristoV300
The northwood's architecture maxes out at 3.6 regardless of the size of the core.

I'm no CPU engineer, but doesn't a die shrink almost always bring increased clock speeds? What about AXP .18 micron -> .13? T-Bred and Barton were nothing more than die shrinks (with very minor CPU alterations compared to Prescott) and they've now added about 1 GHz over the Palomino core.

Based on that assumption and the difficulty Prescott has given Intel, I think that a 90n Northwood core couldn't possibly have done worse for them.

Agreed :thumbsup:

Agreed again. northwoods go to about 4 GHz on good water systems all the time. you can get 3.6 on the stock cooler. and yes you can always make it a little faster just by shrinking the die. i think with some more revisions, fixing the thermal issues, and getting 90 nm working, intel should be able to get decently past 4 with the prescott. the big one there is the thermal issues. if they cant stop all the leakage current netburst is probly gonna die after the prescott and be replaced by dothan.

 

[LTC8K6]

What problems with Prescott?

 

[Vette73]

Originally posted by: LTC8K6
What problems with Prescott?

Joking right?


They run hotter, perform even less then a northwood at the same mhz, etc...

 

[clarkey01]

plus i hear yeilds @ higher speed grades are crap, but from 2.4 to 3.2 Ghz there getting better, notice how there are hardly any 3.4/3.6Ghz Presshots about ?

 

[clarkey01]

plus i hear yeilds @ higher speed grades are crap, but from 2.4 to 3.2 Ghz there getting better, notice how there are hardly any 3.4/3.6Ghz Presshots about ?

 

[LTC8K6]

They are only a little warmer than Northwoods, nothing a good HSF can't handle, they overclock well, and they perform faster in some areas than Northwoods.

I can see the reason for not paying more for a Prescott, but if they cost the same as a Northwood, I see no reason not to buy the Prescott at all.

There certainly aren't a lot of problems with Prescott at all.
In fact, there are none other than trying to justify it over a Northwood if it costs more.

Besides, Northwoods and S478 are pretty much done anyway.

 

[Alkaline5]

Originally posted by: LTC8K6
They are only a little warmer than Northwoods, nothing a good HSF can't handle, they overclock well, and they perform faster in some areas than Northwoods.

I can see the reason for not paying more for a Prescott, but if they cost the same as a Northwood, I see no reason not to buy the Prescott at all.

There certainly aren't a lot of problems with Prescott at all.
In fact, there are none other than trying to justify it over a Northwood if it costs more.

Besides, Northwoods and S478 are pretty much done anyway.

Very true. IMO the biggest remaining shortcoming of Prescott is with clock-ramping. The Northwood finished at 3.4GHz, which is the highest clock speed that Prescott is currently available at. (I know 3.6GHz has been announced, but it wasn't available for purchase on newegg last time I checked.) Since they're ditching the architecture next year anyway, they could have just scrapped Prescott and gone with a die-shrunk/optimized Northwood and probably already been at 3.8GHz by now. Instead they made quite a few large changes and it's basically gotten them nothing except delays.

 

[bunnyfubbles]

I believe the early reviews showed Prescotts pulling ahead of the Northwoods on a clock for clock level at around 3.6GHz. Right now Neither chip has a huge advantage. The extra L2 cache hasn't proven super useful until recently with new games such as Doom 3. But it is true, Prescott was supposed to bring life to the P4 just as the Northwood did after the relatively horrid Whilamette, right now A64s become more and more tempting every day.

 

[jiffylube1024]

Originally posted by: LTC8K6
They are only a little warmer than Northwoods, nothing a good HSF can't handle, they overclock well, and they perform faster in some areas than Northwoods.

I can see the reason for not paying more for a Prescott, but if they cost the same as a Northwood, I see no reason not to buy the Prescott at all.

There certainly aren't a lot of problems with Prescott at all.
In fact, there are none other than trying to justify it over a Northwood if it costs more.

Besides, Northwoods and S478 are pretty much done anyway.

You see no reason to buy a same-speed Northwood over a Prescott at the same price point? How about a 50% longer pipeline and clock-for-clock slower performance? How about increased heat and power consumption?

 

[clarkey01]

?You see no reason to buy a same-speed Northwood over a Prescott at the same price point? How about a 50% longer pipeline and clock-for-clock slower performance? How about increased heat and power consumption??

Agreed, there seems to be this common thought that Prescott overtake northwood @ around 3.6 Ghz, toms showed this isn?t the case, and tom being the Intel enthusiast that he is means that?s quite a bold statement, Prescott may run a tiny bit faster at around 3.6Ghz, but nothing earth shattering.

I also read @ the last minute (early 03) that they increased the pipeline stages to 31, it would have been nice to see Prescott with say a 21 stage (same as northwood).