Which chip will reach market first?

With Hector Ruis saying Barceona will ship in "late summer" - lets call it September for sake of argument, this is getting quite close to the Late 2007 timeline for Penryn (the 45 nano shrink of Conroe/Woodcrest/Merom/etc)

Given that Intel has shown and demoed working 45 nano silicon, pulling Penrym in to the same time frame as Barcelona is not out of the rhelm of speculation.

I think it is unlikely, because while the design is clearly almost good to go, the logistics of ramping a new process is a tough road to go.

What do you think?

Rational arguments only please, evidence, links, that sort of stuff.

Maybe we can have a Baron free thread? 8O 8O 8O

From a completely non technical perspective, I would have to agree.

In terms of cost, for Intel to begin shipping 45nm before AMD gets its competative 65nm out the door is lost profit. Intel has already sunk its R&D costs into 65nm, and as long as their 65nm Uarch is out performing AMDs, they can continue to profit and attempt to regain market share. To release 45nm prior to AMD releasing K10 (except by a week or 2 for premptive PR) would be throwing away potential profit. Addtionally unless Intel is trying for a seriuos technical image PR coo, the more time they spen with Penryn in the lad, the more time they have to detect/minimize any flaws in the design/manufacturing process. So there just doesnt seem to be any great pressure to get 45nm out the door, unless AMD has been stringing everyone along with current availability dates and starts filling the channel/OEMs with K10 in the very near future vs 2008

There is another side though, and that is die space. Intel will be able to increase yields/wafer and thus profitabilty at the smaller node. With the trend towards 4 and possibly 8 cores, increasing yields per wafer becomes increasingly attractive, but since it is unlikely that Intel will cut dual cores from its line up anytime soon, the impetus to transition to 45 nm is not as strong as it could be on that issue

I agree. I would not be suprised (though I doubt we would ever know for sure) if Intel had Penryn ready to go by end of Q2/early Q3 but held it back until K10 was released. Keep the crown, maximize profits at 65nm, and try to steal some or AMDs K10 release thunder. Just a thought with absolutely no evidence to support it. Other than Intel striving to clean the egg of its face after 3 years in AMDs performance shadow. And netburst.

That certainly seemed to be the game they were playing with the release prices of C2D after AMD announced its price cuts for pre C2D release. As well as the release of C2Q.

Ummmm, the code name is not Penrym put Penryn.

I've mentally come up with a couple of scenarios...
1. Intel waits for AMD to launch, then releases ES chips Works for them, IFF they have the goods to beat AMD's benches, plus allows time to start emptying the supply chain of the existing c2d stock.

2. Intel steals the thunder from AMD's launch, and releases ES chips several weeks BEFORE the AMD launch.

Either case, this is all playing silly buggers with my upgrade plans...

What nails? Nothing left to bite. Right now, I know what case I'll get. Probably what RAM, and Vista ultimate for OS.

MB? either sli or xfire...
GPU? Patiently waiting.
CPU? Patiently waiting.
PSU? Depends on all of the above.

Think that I'll just get an opty 180 for the duration...

OK, I'm just a wee person. But if I was a buyer for HP or DELL, I'd be losing hair.

I don't think they're producing a lot of for sale parts at D1D (if any). They've been in charge of development for Penryn and the coupled chipset process (remember that Intel designs chipsets for all their chips). I don't believe D1D is in charge of flash dev. D1D has got to be starting the process for 32nm (at the very least moving in new tools) as well, so they're probably pretty packed (you eluded to this).

I think Jack hit the nail on the head (big surprise) with the idea that D1D will produce enough material to whet our appetites (also known as a full AMD launch), but will wait until Fab 32 is ramped to start offering the chips in volume. I have no idea what stepping they're on with Penryn (the demo was the A1, right?), but they have to be close to "launching". Maybe they really are just waiting to defuse Barcelona on the server side, build volume and then release volume when the AMD desktop parts appear. The next 6 months will hopefully be very exciting!

You and me both. I plan a summer upgrade, but have no idea what to look for.

I vote for Penryn being out first. I've become a bit cynical with AMD and wonder if there will be yet another delay. Even without a delay, I'm a bit suspicious that Intel has a working cpu and is planning a release timed to take away AMD's thunder. Doesn't have to be several weeks in advance, just one week would do, supported by benchmarks in Tom's and every other site showing how good it is.

As to your reason number 1, that could even explain why the prices of the present Intel chips keep dropping. Clear out the inventory and have all new stuff for the customers to buy. Just speculation, nothing more.

My money is on Intel waiting until AMD drops Barcelona before they drop Penryn, there is little value to them beating AMD and it will give them a chance to see what they are actually up against price-performance wise before they finalize their prices.

True

Wolverine, the demo of 45nm Penryn was the A0 stepping. My understanding is that the A1 will be released to the tests labs at Intel fairly soon.

I have not heard if they have removed all production of 65nm parts from D1D yet. But until they update that older Fab down in Arizona to produce 65nm processors I'm betting Intel will keep producing parts at D1D.

D1D is Intel's biggest development Fab that is has. I would like to take a tour sometime through Ronlar campus sometime.

They can't be producing 45nm parts in volume yet because they are still in the development and debug phase of the product and process. That means they would keep producing salable product to more fully utilize the fab.

I would not be surprised if Penryn brings a 5-10% gain in IPC over the current Conroe. We already know about the 50% larger L2 cache and 25% faster FSB, but Anandtech estimates that Penryn contains a 23% higher transistor count in the *logic, control, & L1* side: Anandtech's take on Penryn transistor count. Much of this, of course, is due to the new SSE4 instructions but probably not all.

Intel's die shrinks usually bring increases in IPC:

350nm to 280nm:
This was the Pentium Classic to Pentium MMX transition. Along with the overhyped MMX multimedia instructions it brought a significant boost to IPC thanks to a larger L1 cache and other improvements.

350nm to 250nm:
This was the Pentium II Klamath to Deschutes transition. I think this was pretty much a dumb shrink but I remember the 50% faster FSB (100 MHz) made a big difference in performance.

250nm to 180nm:
This was the Pentium III Katmai to Coppermine transition. L2 cache was smaller but was moved on-die, had massively greater bandwidth and set associativity, and ran at full clock. Overall IPC improved significantly over Katmai.

180nm to 130nm:
This was Willamette to Northwood. Double the L2 cache and other minor tweaks made Northwood an average of over 10% faster than Willamette per clock.

130nm to 90nm:
Northwood to Prescott didn't turn out so well. Though L2 doubled yet again, the instruction pipeline was far longer and the ALUs weren't double-pumped anymore. Despite a host of improvements to the core, those two design decisions made Prescott roughly the same in IPC on average as Northwood, while consuming much more energy and dissipating more heat.

However, the notebook Pentium M's Banias to Dothan transition was much better. Improved IPC (thanks to doubled L2) for the same energy usage.

90nm to 65nm:
Dothan to Yonah. Again, a healthy improvement in IPC. Of course, we know this was much more than a dumb shrink.

65nm to 45nm:
Conroe to Penryn. Will history repeat itself?