Originally posted by: BigDH01
If that was a 3.2C Northwood you bought from your company then you got a great deal. It was faster than the Athlon XPs of that time.
I see a problem with yesterday's launch. They launched a chip that is slower than Intel's slowest quad core part and which requires more power. The only thing competitive about their chip is value. This is not the position to be in if you are getting destroyed by the market leader and squeezed into margins you can't maintain to survive. I would also hesitate to call the products launched yesterday as "pre-production." I believe those are the products you will be buying at the store.
What's so great about AMD's backwards compatibility? The upgrade to Athlon 64 from Athlon XP required a new motherboard. And the first A64FXs required socket 940 boards with registered RAM. How long were those around? Oh yeah, then there's the replacement of socket 939 with AM2. Many people around here were upset about that. I had already sold my X2 for a C2D by the time that happened so I didn't care as much. Regardless, don't act as if Intel is the only company to obsolete a motherboard/socket.
I would hardly say the A64 trounced the P4 for 3 years. In case you've forgotten, the P4 remained competitive in many areas. Look up encoding or 3d rendering. It was hardly a clean sweep for the A64. I realize that many views have been distorted by a gaming-only mindset and the poor reputation of the P4, but for some people the P4 was the chip to have. It wasn't until the high clocked 64s and X2s that AMD finally took a definitive lead in just about everything.
I won't address the innovation comment. I think it's quite obvious that if we define aggregate innovation as being able to produce the fastest chip using the least amount of power at a competitive price then there is a clear winner. If you want to define innovative solely by the interconnect, then so be it. I think that's a fairly limited comparison and ignores architectural advances that exist in Core but to each their own. Your argument would also be more germane if the FSB really was a bottleneck. All indications I've seen are that the FSB is not a huge bottleneck. It certainly doesn't stop Core from outperforming Athlon64/Phenom (at least on the desktop).
AMD released the Phenom because it has to try and stop the bleeding. It can't just scrap all those expensive MCM chips it's produced even if they do have errata. They aren't in business to lose money.
Why would I be limited to a Celeron D in a 300-400 dollar system? I can get an Allendale for only slightly less than $20 more than an X2 3800+. And that's only if I opt for the OEM 3800+. If I go retail then the difference is less than $15. Worth it in my opinion. Tack on another 50 dollars or so for a motherboard with integrated graphics and you are well on your way to a dirt cheap Core 2 system, albeit one with crappy graphics performance. (All prices from Newegg)
My first AMD chip was the K6. I've owned just about every iteration of AMD chip after that. It doesn't mean I blindly worship AMD. This launch is certainly lackluster if not a complete failure.
You raise many good points. Netburst was ultimately a failure because Intel ran into severe thermal issues with Prescott, the architecture itself, whilst lacking in IPC, was capable of clocking into the stratosphere, it was simply overcome by curent leakage on the 90nm process. 65nm actually fixed the leakage issues somewhat, people were overclocking to 5GHz+ on air with the latter 65nm P4s, but by then the world had moved on to dual core, and even 65nm PDs were relative ovens compared to the more efficient X2s. Then of course, came C2D, and the rest, as they say, is history...
You are correct in saying that P4 was competitive in rendering and encoding, mainly thanks to it's SSE2 implementation that AMD did not incorporate until their Venice A64s if my memory serves me correctly.
People used to harp on the gaming prowess of A64 compared to P4, and whilst at low resolutions and detail settings it did trounce P4, in 'real world gaming' situations the difference was nowhere near as apparent.
Notice how nowadays, people emphasize the importance of GPU over CPU in the overall gaming experience? 'GPU limited' is a popular term amongst enthusiast gamers nowadays.
The same applied between the P4 and A64, once you ran games at higher details and resolutions, the limiting factor was predominantly the GPU, not the CPU. I tried pointing this out way back in 2003, to be shot down by fanboys saying 'STFU, look at the benchmarks, A64 pwns P4 at gaming!'.
In terms of upgrade paths, both AMD and Intel have been guilty of killing off upgrade paths on their platforms.
Since 2003, AMD has released S940, S754, S939, and AM2 (and now AM2+ obviously) for the desktop. That's roughly 1 new socket every year, and the one that was most disappointing was the discontinuation of S939, as it left a lot of enthusiasts out in the cold.
Intel, whilst having stayed on S775 for the past few years, tends to introduce new VRM requirements on each new generation of processor, and kept introducing new chipsets like it was going out of fashion or something. Sometimes the new chips would work on older boards, sometimes they wouldn't, it was really a crap shoot.
A lot can be argued over the technical merits of the native vs MCM approach, and frankly I would leave more technically savvy people to debate those points, but at the end of the day it's not hard to see who won this round. Intel's MCM quads may not be the most elegant by design, but the fact remains that Intel's slowest quad is faster than AMD's fastest quad, and that's a rather damning fact. Even with the native approach, Phenom couldn't achieve clock for clock parity with a 12 month old C2Q design, let alone the upcoming 45nm versions.
People have to understand the relative strengths and weaknesses of Phenom vs Core 2. K10 is strongest in servers, where it's scalability with increased core/socket count is far superior to the FSB constrained Core 2 Xeons.
However, desktop performance is the strength of Core 2. The bandwith advantages of HT vs FSB really doesn't mean much on typical desktop applications. That is the hard truth. Sure, AMD does show better scaling on heavily threaded applications (see Cinebench scaling vs Core 2), but it's still merely playing a game of catchup, rather than leapfrog. Ultimately, IPC still rules on desktop, and all that extra bandwith means little in real world desktop usage.