Nehalem EX "Beckton" architecture

Holy shit. That's definitely not fair!

When is AMD going to release something new? They've been dragging that Athlon64 X2 shit for pretty long time now, ever since the 2.4C (800Mhz FSB Northwoods w/ HT) came out, from what I remember.
 
RangerSVT said:
Holy shit. That's definitely not fair!

When is AMD going to release something new? They've been dragging that Athlon64 X2 shit for pretty long time now, ever since the 2.4C (800Mhz FSB Northwoods w/ HT) came out, from what I remember.
Click to expand...



I know,its pathetic ! :)

By the way....take a look at the OC thread I started,in the AMD sub foum... juicy stuff..
 
I almost can't believe this... wow! And I was impressed when I saw the QX6850 overclocked to over 5GHz w/liquid nitrogen.
 
mito said:
when is it gonna end?
Click to expand...

I donno, I get a tear in my eye when I think about it. However, life is cycles. So, maybe in a few years AMD (if they are still around) will make a come back.
 
This is basically intel kicking AMD while their down and using a baseball bat, and I love it.
 
I know there's going to be alot of people coming in saying "OMFG U GUYS R TEH L000Serz,,, amD noT GONNA GO ANYWhuRRRR...".

Well I sure as hell hope they don't, keeps competition going and prices dropping.
 
Going to 4 cores this year and 8 cores in 2009. Awesome!! :eek: 8 cores in one die is nice also.

How will be temps be for the 8 cores?
 
SCSI said:
Going to 4 cores this year and 8 cores in 2009. Awesome!! :eek: 8 cores in one die is nice also.

How will be temps be for the 8 cores?
Click to expand...

My guess is the same as todays processors, possibly as high as 150w to as low as 95w (I would say its more likely to be higher, on die memory controller and looking at the diagram, L3 cache is only going to add to the heat output.
 
I'm more interested in Dunnington personally.
That may be a drop-in replacement for many socket 775-owners.
I might skip quadcores altogether and go to six cores right away.
 
RangerSVT said:
Holy shit. That's definitely not fair!
Click to expand...

It's business, it's rarely fair. Intel makes more money in a quarter than AMD sees all year. Intel has also been profitable for the last 20 years, AMD not so much. I think Intel was caught off guard by AMD when they released the Athlon 64. I don't think Intel will allow something like that to happen again anytime soon.
 
here's my wild prediction, so i can check back in 7 years and see how far off the mark i was:

12nm, 16 cores (standard, up to 64), 4ghz, with an IPC 2x higher than Core 2
 
fluxion said:
here's my wild prediction, so i can check back in 7 years and see how far off the mark i was:

12nm, 16 cores (standard, up to 64), 4ghz, with an IPC 2x higher than Core 2
Click to expand...

4Ghz only?

Make it 6Ghz
 
RangerSVT said:
I know there's going to be alot of people coming in saying "OMFG U GUYS R TEH L000Serz,,, amD noT GONNA GO ANYWhuRRRR...".

Well I sure as hell hope they don't, keeps competition going and prices dropping.
Click to expand...


That I fully agree with. :)
 
I agree as well. But then again would Intel take advantage of us if their was ever no competition?
 
Nemik8800 said:
I agree as well. But then again would Intel take advantage of us if their was ever no competition?
Click to expand...

I've been using PCs since the late 80s, long before AMD even made its own line of CPUs, and certainly long before AMD actually was any competition (everything before the Athlon was basically low-budget stuff, which I never bought anyway).
It wasn't that bad actually. In fact, it's back to that situation right now. The faster dualcores and the quadcores from Intel are completely unthreatened by any competition at the moment. For well over a year AMD has not been able to deliver anything faster than my current system. So AMD simply doesn't matter to me at the moment. They're flying below the radar, as they always have for me, except for the time when I bought an Athlon.
I don't think it's going to matter much when AMD drops out of the picture completely. It'd be just like old times again. And they weren't that bad (although I still prefered Motorola CPUs at the time ofcourse, but they weren't direct competitors).

I just wonder how many of you panicking people have ever experienced the times before AMD?
 
Scali2 said:
I've been using PCs since the late 80s, long before AMD even made its own line of CPUs, and certainly long before AMD actually was any competition (everything before the Athlon was basically low-budget stuff, which I never bought anyway).
It wasn't that bad actually. In fact, it's back to that situation right now. The faster dualcores and the quadcores from Intel are completely unthreatened by any competition at the moment. For well over a year AMD has not been able to deliver anything faster than my current system. So AMD simply doesn't matter to me at the moment. They're flying below the radar, as they always have for me, except for the time when I bought an Athlon.
I don't think it's going to matter much when AMD drops out of the picture completely. It'd be just like old times again. And they weren't that bad (although I still prefered Motorola CPUs at the time ofcourse, but they weren't direct competitors).

I just wonder how many of you panicking people have ever experienced the times before AMD?
Click to expand...

dont really remember the pricing so i cant really vouch for them being competitive in pricing in their days of unchallenged dominance. but you look at architectures like Netburst, you see a situation where marketing (more ghz = more sales) prevails over performance (ignoring the engineers who realized that a complete redesign was needed and inevitable.....(i saw a talk one of their engineers made about his time there working on the Pentium 4 and it's would-be successor)).

they will undoubtedly take advantage price-wise as well, any company would, and if they didnt they wouldnt be serving their shareholders very well. the question is by how much? you cant cite the current situation for comparison, as Intel is still heavilly investing in reclaiming the market and asserting their absolute dominance, which include low-pricing and an excellent technological road-map for the short-term. it's only after they've done that before we'll start to see how they use their position with regard to pricing.
 
fluxion said:
but you look at architectures like Netburst, you see a situation where marketing (more ghz = more sales) prevails over performance (ignoring the engineers who realized that a complete redesign was needed and inevitable.....(i saw a talk one of their engineers made about his time there working on the Pentium 4 and it's would-be successor)).
Click to expand...

Yea, hindsight is always 20-20.
I don't buy this "Netburst is all marketing"-thing. Firstly, what would Intel be marketing? You have to realize that Netburst started years before its introduction. They had probably already outlined the new architecture before the first Pentium Pros were even on the market. Long before Intel ever had any kind of competition whatsoever. So why would they be marketing GHz? They had the fastest x86 processors by default, there was no competition to speak of. Intel could in no way have foreseen AMD's success and AMD beating Intel to the 1 GHz barrier with the Athlon, at the time they chose to make Netburst a high clockspeed architecture. When Intel made this decision, they were the fastest x86-maker, and have always had the highest clockspeeds by far.

Secondly, I think you are selling the Netburst architecture short if you say it's about marketing, not performance. Sure, GHz are a good marketing tool, but if you look beyond that, you can see that Netburst also delivered incredible performance, especially during the Northwood era. At this time, the Pentium 4 outperformed the Athlon XP in all areas, and by a considerable margin in some.
Also, look at IBM. They are now going for high clockspeed aswell. You think IBM is just doing that for marketing? You think they're making the same mistake as Intel?
No, high clockspeeds are an excellent tool for high performance, and Intel (and most other chip designers) has historically always designed new architectures that would considerably clock higher than their predecessors.
Just because the die-shrink from 130nm to 90nm backfired on the Netburst architecture isn't a good enough reason to claim it was all about marketing. Netburst was also a very clever and sophisticated architecture, of which many features will live on in many x86 generations... But some of it was just a bridge too far with the technology of the time. Netburst was originally designed to operate at speeds in excess of 5 GHz, and only then will it reach its optimal balance of IPC and clockspeed for the best efficiency. But as it turned out, their production processes limited the CPU to sub-4 GHz speeds and more power dissipation than originally anticipated.
Don't be surprised when Nehalem or architectures even further into the future will resemble Netburst more closely again.
Larrabee, the massively parallel x86-like processor, allegedly to be used as a high-end GPU, has been demonstrated at over 6 GHz. This also reeks of Netburst technology.

Don't sell Intel short here. I'm sure that if any other company had the resources at the time, they would have tried a project like that. At the time it seemed to make perfect sense, and at first, it actually did. But on then 90 and 65nm processes that followed, it appeared that a different strategy was far more effective. Don't forget that architectures are very much a sign of the times. If anything, Netburst seemed too far ahead of its time. Many of the ideas in Netburst make perfect sense in theory, we just cannot produce them in practice... yet? Therefore a less ambitious architecture such as the Core2 or Athlon64 is a better choice today.

Now, I don't believe the marketing department was the 'mastermind' behind the Netburst architecture. I am quite sure that this is what the team of experienced chip designers at Intel considered to be the future, after the Pentium Pro at the time (remember, this was early to mid 90s).
I do believe that the marketing department had a key role in the fact that the Pentium 4 is the longest running architecture (and probably most successful) in the history of Intel.
What the engineer may be referring to is the fact that the marketing department managed to keep Netburst on the market and selling well, long after it had lost the performance crown. Perhaps the engineers wanted to release Core onto the desktop market sooner, because they knew they could beat AMD with that. But instead, Intel could take its time designing a native dualcore solution that was far better: the Core2. In the meantime they used the Pentium 4 to refine their 65 nm process. The result? We all know that... Intel didn't just take back the performance crown from AMD... No, they degraded AMD to a bargain basement brand, almost overnight, with the introduction of the Core2 a year ago. AMD still hasn't recovered from that one. Now this is very much because of the marketing department.

Besides, it doesn't make sense... Intel was always very successful in designing and marketing new CPUs... Did they suddenly all turn into idiots when they thought of Netburst? It just doesn't make sense, with such a large company with such an excellent track-record, and so many skilled and experienced people working for them. And how do you explain Core2 then? They suddenly got smart again?
No, I don't buy it.
I think Intel was as smart as always, they just had a set-back that messed up their plans. It happens.
The Pentium Pro was all about high clockspeeds aswell, delivering little or no IPC gain over the classic Pentium in many cases (or even worse in some), but while the Pentium ranged from 60 to 233 MHz, the Pentium Pro made an incredible leap from 150 MHz to about 1400 MHz (even more if you include Pentium M and Core in the Pentium Pro family). All about clockspeed. And where was AMD back then? Sure, their Athlon was successful, but that was partly because it was largely a clone of the Pentium Pro architecture, and AMD borrowing technology from the legendary DEC Alpha architecture. The Athlon was a architecture, but hardly surprising when you combine elements from two earlier architectures that are some of the best in the industry. Now that both the Pentium Pro and the Alpha have run their course... What rabbit is AMD going to pull out of its head this time, to beat Intel?
Because AMD is still trying to stretch the Athlon further, and it no longer wants to cooperate. The Athlon has become AMD's own little Netburst, which they cannot scale to the clockspeeds they intend, which they cannot get down to the power dissipation envelope they intend, and they have to keep the public happy with hollow marketing such as 'native quadcore'. So it's time for AMD to close their 'Netburst' down, and introduce their own 'Conroe'.
 
Good post Scali, much more thought out then what the official line on netburst has become, and I totally agree with it.
 
Comment: Below is just my thoughts, a lot depends on where you were standing at the time all this happened and I was certainly not standing inside Intel. So just my take on it, not saying anything or anyone is wrong etc., just my observations from my viewpoint at the time. So all worthless speculation then.


Besides, it doesn't make sense... Intel was always very successful in designing and marketing new CPUs... Did they suddenly all turn into idiots when they thought of Netburst? It just doesn't make sense, with such a large company with such an excellent track-record, and so many skilled and experienced people working for them. And how do you explain Core2 then? They suddenly got smart again?
Click to expand...

Having worked as a design engineer and in IT for 20 years in a fortune 100 telecom company I can attest to the engineering graveyard of good ideas that did not quite make it for one reason or another or where put to market after tons of money spent to go nowhere. To anoint Intel with some "they cant make a mistake" mantel of infallibility is unreasonable. They were just large enough in a corporate and marketing share sense to weather the mis-step. Outside of the insignificantly small "enthusiast" market, and stepping into the the "big corporate, machine on every desk, office drone computer market, none of netburst P4 shortcomings where even noticed except for the hardcore IT guys and we always had ways around corporate IT's "list of approved hardware" to build our own workstations and frequently did. Like the IBM saying, no one ever got fired for specing in a Intel cpu." Core 2 is built on the Intel Mobile CPU that was widely seen as a better path forward in design, it just took Intel a while to realize it. Also do not underestimate the politics between different design teams and marketing when clock speed was seen as "THE selling point". I do not think there are many idiots at Intel but there are most definately people with agendas and my experience tells me that seldom is it for "for the good of the company and its customers first" exclusively. Other concerns are always present.

That said, Intel has never really made a "bad" cpu. They all have their place and work pretty much as intended and with the amazing progress made since the 8088 it is difficulit to throw rocks at them. All I am saying is that it is very easy to accidently go down a path that is not optimum when you are out in front leading the majority of the time. And when you are very big, the inertia makes it harder to get back on the best path.

er, the IBM processor is a RISC architecture, clock speed IS king in that context.

their production processes limited the CPU to sub-4 GHz speeds and more power dissipation than originally anticipated.
Click to expand...

yep yep could not agree with you more, however I do feel they could have seen that coming a bit sooner. We (speaking of our internal chip fab expertise) did. Perhaps Intel realized it too, but that brings me back to the "its difficult to be nimble when you are big" point.

er, crap, what was the question LOL.


Oh SHIT, the solar panel on the space station is damaged, now that is important, ignore all the above if you wish as it is just my view of history that has many possbile views depending on where you were standing when it happened. Probally not worth the time it took to type it as it is all water under the bridge now.
 
Fuser said:
Something I think a lot of people miss that really helps the Core architecture was the amazing work in branch prediction done in the Netburst architecture. Oh well it wasn't a total loss :)

http://arstechnica.com/articles/paedia/cpu/p4andg4e.ars/4
Click to expand...

Yea, some things from Pentium 4 that were and are great technologies:
- Branch prediction
- SSE2/3
- Trace cache
- HyperThreading
- Double-pumped ALUs

SSE2 and SSE3 are still what give Intel the edge in video encoding today... AMD supports it, but they couldn't touch the performance of the Pentium 4 at the time, and still can't touch the performance of the Core2 today in these tasks.

Intel has announced that HyperThreading will return on the Nehalem core. We'll be getting 8 cores, capable of running 16, or even 32 threads?

I've not yet heard anything about trace cache, but I always considered this a great idea. It makes perfect sense to decode the x86 instructions only once. Most of the time is spent inside loops anyway, so you keep running the same instructions. When it's predecoded in cache, you save yourself the trouble of decoding it over and over again, and you can put extra hints for memory and branch prediction inside the cache, etc.
Somehow I think this is going to make a comeback aswell.

The double-pumped ALU idea... I guess they're using something like this to get the insane clockspeeds in their Larrabee chip. The idea of having some kind of 'early out' on simple instructions makes a lot of sense. I can see this returning aswell in a future Intel architecture.

So I basically see a Pentium 4-like architecture in the future, but with a pipeline closer to the current Core2 architectures, and more ALUs instead of faster ALUs.
We'll have to see if Nehalem is what I think it is. No details out yet.
 
Has anyone done the calculations yet?
In the case that 45nm is overclockable up to around 4.6GHz:
4.6x8= The equvilence of a 36.8 GHz Pentium D
Now with skulltrail it brings it up to 73GHz equvilence.
Can I get a :D?
 
Nemik8800 said:
Has anyone done the calculations yet?
In the case that 45nm is overclockable up to around 4.6GHz:
4.6x8= The equvilence of a 36.8 GHz Pentium D
Now with skulltrail it brings it up to 73GHz equvilence.
Can I get a :D?
Click to expand...



The ^ above looks like a post Dan D would frown on.... :D

I wonder how the 24 way set associativity in the Penryn's L2 will help over the Core2's ?
 
Manny Calavera said:
The ^ above looks like a post Dan D would frown on.... :D

I wonder how the 24 way set associativity in the Penryn's L2 will help over the Core2's ?
Click to expand...

You are absolutely right. That's simply NOT how multi-core processing should be thought of nor is it how it works. Performance does not scale that way as cores are added.

Nemik8800 said:
Has anyone done the calculations yet?
In the case that 45nm is overclockable up to around 4.6GHz:
4.6x8= The equvilence of a 36.8 GHz Pentium D
Now with skulltrail it brings it up to 73GHz equvilence.
Can I get a :D?
Click to expand...

No.
 
Anyone know if Beckton will be a quad socket offering upon release, or will Intel also make it available for dual socket systems and desktops? I can't find anything conclusive on this matter and some info I actually found indicated it's designated for MP architectures.
 
You must log in or register to reply here.
Back
Top