Affordable Nehalem out to clobber 45nm AMD
THIS WEEK, the "Alpha Male" of AMD, Captain Dirk, took over the troubled ship.
His Alpha processor experience did help create some of the fastest X86 CPUs ever with the first Athlon64 and Opteron offerings, but this time he's not starting from a pole position - or even the more favourable position he could have held if AMD had kept the Alpha going alongside the X86 in socket-compatible line-up that would fight Intel across the board far more efficiently.
There are web rumours, such at those at Digitimes, that even the extreme Bloomfield might end up at less than US$1000, less even than the current extreme Yorkfields? And that's a badly diminished US$ at that. At this year's currency rates, Euroasians might find these upcoming Nehalems very affordable actually. Not to mention the still-decent 2.66GHz Bloomfield at less than third that price? The expected yields must be darn good if these rumours are true.
I don't believe Intel would be that quick to offer a brand new premium product at a price that's lower than they could afford charging. But, again, the "extreme desktop" CPU category was always in the $1K range, except for the past year's aberration where the QX9770 went quite a bit above that - simply because there was, and still is, no competition anywhere near it.
Of course, such a return to the old ways would be good news: you could expect corresponding price adjustments to the existing Penryn generation prior to the Nehalem unveiling, making these otherwise excellent CPUs even more attractive. As we said before, the one category where there's least reason to move from Penryn to Nehalem is, well, uniprocessor desktops. The FSB is a limitation only in workstation and supercomputing apps as of now.
What does it mean for AMD and its new CEO, who still has to share the control with his executive chairman? Well, one could look at the rumoured attractive Nehalem pricing as a possible knock down, if not knockout, blow to AMD CPU business.
Knock down? If AMD does deliver enough 45nm quad-cores this year - enough means probably covering all the press, OEMs, channel demo systems and select key customers and gaming sites, plus a few thousand actually sold - they'd be seen as making a sort of minor comeback. However, the performance gulf would still be kept wide "advantage Intel" by the new Nehalems at decent prices, not to mention current Penryns for even less dosh.
Knock out? Well, if AMD doesn't deliver the 45nm at all this year, and the new LGA1366 socket end up this affordable even before those really cheap LGA1160 Lynnfield and other 2009 Nehalems start appearing, AMD might really lose the high-end desktop completely. Same with the workstations with the Gainestown DP Nehalem addressing the competitive position in a similar fasion
For any comeback, AMD will then have to work triple hard - proving itself back on performance, features and pricing all at the same time, without the " underdog" benefit they had before. Simply, the situation where (now former) CEO takes huge bonuses when the company has its largest losses in history has taken away that compassionate feeling in many.
On the other hand, if AMD does the 45nm right, and manages to ship decent amounts this year, with good pricing to boot - i.e. below US$ 300 range for a 2.8 GHz quad core - it'd have a chance to be seen as solidly back in the mainstream despite the Nehalem onslaught. Of course, a strong marketing campaign would have to trumpet that new-found confidence to the world.
We can talk then of a nice mainstream battle of 2.66++ GHz generation: Yorkfield, Bloomfield, and Phenom, all in the 2.6 to 2.8 GHz band at around US $250 to 300. What a sweet spread to pick from, affordable almost to anyone. And, you need not compromise on your fancy GPU to afford a decent quad-core CPU here. Makes the CPU-GPU discussion moot, doesn't it?
Share this:
ShareBusiness Intelligence Software
Customer Relationship Management (CRM)
Enterprise Accounting Software
Enterprise Resource Planning (ERP)
Enterprise Systems Management
AMD unlocking their CPUs, and not just the "Black" editions, would go a long way to having their processors taken up by enthusiasts. Their current lineup doesn't take to overclocking exceptionally well, but perhaps their next generation of 45nm CPUs might.

Not that I will buy one even if they do. I have an e8400 OC'd to 3.6Ghz, which should last me for the next 3-4 years (with other component upgrades). :)

Cheers,
John
Nahalem is 4 Cores with 8 threads, So its Everything reliable in One Package Today & affordable. Yet AMD Has Secret Weapon, As reported by Msr. Magee, when still had strength to type. Heres idea puppy:

Torrenza is an initiative announced by AMD to improve support for the integration of specialized co-processors in systems based on AMD Opteron microprocessors. Torrenza does not refer to a specific product or specific technology, though the primary focus is on the integration of coprocessor devices directly connected to the Opteron processors' HyperTransport links, and other co-processors connected via PCI Express. (direct connect)

Yeah,theres idea that Broke: Barcelona by using such faster 2 way L1 crossbar about its town, whole gizmo could go from 20-30 complete changes/second to thousands, Only rest of system wasn't ready, so cut off all extra operations & just pushed signal thru,

NOT Bad in Lab,in field seemed to become cutback to: uni directional in last specs, yet its still tremendously powerful idea, especially with libraries of threads seperate or even interacting as developement. 
Just slow down internal return info to more harmonious clip & by having three out crossbars from L3, maybe to i.e. 200,400 or 6/800 levels internal sub cores & seek best that works, something got to run Mighty Fast, if thats case.
Like Giant RattleSnake in Cage of Mice, theres gonna be some interactions.

Now if I where intrested in tech dept, I'd go Dunnington, as few extra grand isn't much for Next Generation that will fight AMD Machine. However, even with many transistors & Faste as Blazes simpler transistors behind that in developement, its still might be early:Old Mother Hubbards Shoe compared to perfected two way L1 crossbar keeping all libraries instantly updated.
At Any Rate Bloomfields 6 memory slots give lots of birthe to todays machines. AMD plans 8 memory slots coming with those radical new 8core, Instant Ultie, Ultienizers.Nothing Could Be Finer Than Spend upcoming Birthday Summers In Carolina In NAVY.
STeWie drashek

IIRC at some point in time Nehalem was to include an integrated GPU on chip, perhaps not on die though.

Is this still what Nehalem is supposed to be?
Last Night, After Digital Acceptance Corps had gone asleapie, Intel got out word Nahalem,Bloomfield,2.66=$284. Over? $113 Price Cut, it goes with Last of Line & Something To SELL.
Iagree as This is supposed to be Pinacle, UNLOCK MULTIPLIER & CORE. Man Whomn We Await truth From, Mike Magee, Wheres Your 6.4 Ghz/s Now, IN MAINBOARD?

hahahahaha.Power7/CELL. Dunnington+300001 cores Take it TO TEN. new slogan:TAKE IT TO TEN.

Thomas S von Drashek
I predict it will go back to the way it was in the old days. For you whippersnappers that don't remember: Intel was Porsche,while AMD was a Chevy. If you had the cash you went Intel for the Horsepower,if you wanted a PC on the cheap you went AMD because it was "good enough". Now that Intel has gone with Core tech the ability to beat them in a horsepower race is going to be slim. 

AMD,if it can survive the ATI merger,will probably end up being the "cheapie" CPU that is "good enough" just like before. The only question is whether Intel will price them right out of the market with Core tech, as Intel CPUs can be had for pretty damned cheap right now and they overclock much better than AMD. The only thing they have going for them right now is ATI but I doubt that is enough to sustain them. But that is my take on it,YMMV
I wonder, did the Inquirer missed the leaked Deneb benchmarks? Did the Inquirer missed the leaked Nehalem/Bloomfield benchmarks? I do question their "professionalism" often.

Here's a breakdown... 

Deneb 3.4GHz OC SuperPi 1M at a paltry 20 seconds....

Bloomfield/Nehalem 2.4GHz SuperPi 1M at 17 seconds!!

Bloomfield/Nehalem 2.66GHz SuperPi 1M at 15 seconds!!!

Even Yorkfields can do better than Deneb 3.4GHz with lower clocks on this simple test.

AMD's shiny new 45nm has not closed the gap at all, its getting wider... 

Higher clocks simply cannot do it anymore! AMD really needs a 4-issue core instead of the leagcy 3-issue one currently in K10 (and K8). Faster cache too!

Whoever uses SuperPi to benchmark CPUs is an idiot.
They both should be equal in performance. 
Both will be on 45nm.
Both will have integrated memory controller. 
Both will have similar clock frequency range.

Whats the big deal?

AMD Barcelona scales better than Intel Quad-core on multi-core benchmarks.
Lots of folders use superpi to 'benchmark' before the chips come out, atleast for the smaller WUs. Obviously after the chips come out they use folding to benchmark. :-)