Last message on previous page:
The problem is when AMD bought ATI they were on top. Amd was making money. They had an architectural advantage to Intel (Netburst). Amd was very short sighted believeing that this will still be the case allowing them to fund R&D for the future products like fusion.

If you go back to the original articles back when Amd bought ATI they did say that they weren't interested in the discreet gpu market. They just wanted ATI's IP for use in future products. The problem is that they assumed financially it didn't really matter if ATI wasn't bringing in that much revenue because Amd only had to float itself long enough for some of the ATI technology to filter into AMD's product line and become the Spider platform we see today.

However AMD ran into 3 problems.

1. The Core architecture blew Amd's K8 away. It was faster and more efficient. Efficiency probably being the worst factor seeing how much Amd used to grill Intel for Netburst being a radiator. Then all of a sudden Amd's cpu's didn't look that impressive.

2. This isn't fact its just my opinion but I think Amd have just confirmed it. ATI's IP wasn't as impressive as it was made out to be. Amd has also just said that they overpaid for ATI which leads me to believe this to be correct. Ati's chipsets have always been inferior to the competion. There gpu's have always been inovative but they havn't ever managed to be the market leaders. When Amd bought ATI they thought they were getting a simalarly inovative company. What they got was a good discrete gpu maker and absolutly nothing else.

3. The K10 architecture is slow, inefficient, very complicated and expensive to produce. And not only this its almost 12 months late. Not really much else to say about that.

Heres an article on the AMD/ATI merger.
http://www.tomshardware.co.uk/will [...] 21354.html

You may be on to something here -- I'm too lazy to look it up to understand what is issue is -- I take it they are segmenting the mainstream/enthusiast via the IMC? I'm curious to understand what the main issue is from your point of view?

In some ways intel is taking a pretty big risk with Nehalem -- the modularity they are touting (do you *know* how many different flavors of NHM are being punted around?) could be it's downfall. You can only segment the market so much... and with so many versions, it's easier to not predict the product mix correctly.

On the face of it, it seems a smart move to relegate the budget systems to the FSB and the enthusiast systems to the IMC (and, of course, the servers) for cost-effectiveness. I wonder what the other ramifications are -- and if they were thought through. (lack of upgrade paths, confusing product distributions, overclocking issues...)

I think AMDs main problem was that they didnt see it coming! what could beat ta athlon x2 4800 back in the day? a pentium D? i dont think anything. They shoulda tried to invent a new architecture while they were ahead, instead of fiddling around with the cache, and the clock speeds. AMD could beat intel when it was a failure (netburst,) now intel something going...if intel was smart then they should plan ahead and find a new architecture b4 AMD gets something.

I'm not so sure. I was looking at old reviews recently because I'm thinking of buying a dirt-cheap PC from the local computer store with an older AMD CPU to use as a server for our house, and the reviewers were expecting the Core chips to be as far behind AMD as the Pentium-4s; so I think Intel surprised many people, not just AMD.

If that was true, ATI would have gone bust long ago... a decade back there were something like twenty GPU manufacturers, most are long gone. In reality, ATI have tended to be behind Nvidia, but they were ahead for a significant period in the last few years.

The 9x00 series utterly destroyed the Geforce FX, and held the lead for a substantial period of time. Maybe ATI just got lucky, but they certainly led the GPU market for gamers at that point... only nvidia fanboys and people who knew no better were buying FX cards.

As an example...

If X38s were a different socket than P35. X38 socket only got Extreme Edition releases while P35 got the "mainstream" releases. I could see it happening just because of what is going on with Skulltrail. Dual socket, expensive motherboard, and only going to get Extreme Edition CPUs.

Limits your upgrade path when you get tied into a performance bracket. The way it is now say I have $350 bucks. I can buy an X38, drop an E2140 in it for the time being until I can upgrade to something better. How will it be with Nehalem?

It's just going to come down to the Implementation used by Intel. I do know that the highend will be the only one with Quick Path Interconnect. The mainstream will feature 1x16 or 2x8 PCIe Gen 2.0.

That's why I'd like to see AMD come back, that decision by Intel I see as a chance for AMD to gain some market share in the enthusiast market.

Edit: Not to mention the fact that it's bad enough when there are 2-3 Sockets... Now we will have: LGA775, LGA771, LGA1366, LGA1160, and LGA1567.

Oh, I see. Yes, the multiple configurations is a bit disheartening, but the IMC version, as exit2dos mentioned, will be on servers only, first. The desktop variant will be FSB based, still. At least, that's what I understand.
When Nehalem does go IMC for desktop, I can understand the frustration of not knowing which configuration to use.

That's is what I understood as well, although, I cannot find the article that made mention of that. I do remember reading about the first desktop Nehalems being FSB controlled still, though.

Intel's "pricing" war is a bit interesting when holding a monoply state. But Intel is thinking just like AMD back in the day, crush the competition and crank out performance and price. Intel is moving full blast and LOSING money on some cpus (And they can still pull a huge profit for other things). Intel is in the state of mind to take out AMD and it's working great. Caught AMD trying to act like the OLD Intel and with their pants down. I'm surprised Intel isn't slowing down, they got a HUGE lead now. Looks like they will keep going strong until AMD is useless, and "buying" them (AMD) would be worthless. Great move for Intel to be the ONLY supplier. But will obviously come back and bite us consumers.
It's not looking good, almost looking that it might be the end of cpu and video card "wars".

I heard that the Desktop will be quadcore - which wouldn't need an IMC, but I can't find any reference as to whether or not the DT will also have HyperThreading. However, at the last IDF, Intel mentioned integrating graphics on the die - which would probably necessitate an IMC - so it does seem very confusing. At least for now. I'm sure things will fall in place as Intel makes firm product announcements.

From what I've seen/heard and read... All Nehalems will have Hyperthreading. The "on-die" graphics won't actually be on die. The native quad will sit next to a graphics die and just be "glued" together on the package.

Did you just use "Native" and "Glued" in the same sentence? - Blasphemy!!!

Anyways, you're probably right as Intel used the words "in the socket" not "on the die" to describe the gpu.

These were the main things I wanted to respond to in your reply. So here goes...

When AMD purchased ATI, ATI was in a really good hold with the X1900XTX and the upcoming X1950XTX. No one expected NVIDIA to launch the 8800GTX and utterly destroy ATI for the performance crown. But even than ATI was still a good contender and was still selling X1900 series cards because they were quite a bit cheaper. It wasn't until a 6 month barrier hit and the public realized that no ATI counter-product had been released, that so many hopped onto the 8800 series bandwagon; and for good reason, the 8800GTS 320MB was released, which put great performance at a really great price point. This was one of the major flaws seen from ATI within the past 18 months.

Second thing to note is that Intel was releasing a LOT of Engineering Samples to people and a lot of sites, and as I recall one of the best and most popular of these sites was Xbit-Labs. Not only this but a sample was provided to Anandtech, that sample, like all the other samples, were of the Core 2 Duo E6600 processor. They benched the E6600 against AMD's finest, the Athlon 64 FX-62. At the time it was $1,000+ and held the performance crown against the Pentium D's, both Presler and Smithfield alike. But once word got around that a suspected $300 processor from Intel could trump a $1,000 processor from AMD, people started purchasing 975X BadAxe boards like crazy to prep for the Conroe onslaught.

Basically what I'm getting at is this; AMD was aware of how the Core 2 Duo architecture could compare to it's K8 architecture. This means that they really have no reason or excuse for K10 not performing well. They KNEW what Core 2 Duo could do, yet here we sit. ATI knew NVIDIA had a better product, yet didn't release anything to compete against it for over 6 months. This is unacceptable in my opinion.

AMD knew Core 2 Duo could trump K8, ATI knew they needed an answer to the 8800 series, so WHY are they behind so much? They had more than enough time to prepare! I guess they figured that the AM2 launch would be great or something, I do not know. What I speculate is that if AMD had designed K10 to NOT use an L3 cache and use the traditional L2 cache, than they would be in a much better spot, as their processors would probably be performing much better.

Hmmm... think about it... that implies you can get a quad core + GPU (with one chipset, not aimed at enthusiast, since they'll go discrete) OR 8-core w/o GPU. (another chipset, aimed at enthusiast)

I can see where this bifrucation would get maddening....

But, I don't think the situation will be as bad as you imagine. I can see a "high-perfiormance" socket (with IMC, QPI and the extereme chips + other higher end chips) and a mainstream (read: low) performance socket (with FSB and lower clockspeeds) This will effectively mean that the enthusiats looking for getting the el-cheapo chip and O/C the hell out of it are out of luck. Maybe. Let me check to see if there are any roadmaps I can spot.

I found this at ars:

http://arstechnica.com/news.ars/po [...] r-smt.html

They're speculating that it will be a dualcore with a GPU.