Slashdot Mirror
What Went Wrong for AMD's AM2?
An anonymous reader writes "When AM2 was first announced it seemed like it was going to be a guaranteed hit. After all, this platform would be moving the tremendously successful socket 939 into the future with its use of DDR2 memory, a greatly increased memory bandwidth, hardware virtualization, and a number of exciting new CPUs. Despite everything AM2 had going for it, this includes a dedicated enthusiast base and a tremendous amount of pro-AMD spirit at the time, the new platform has largely been dismissed by consumers. The question now is, what happened? How did AMD go from record growth and being the darling of enthusiasts to having a new platform which failed to impress?"
Ah, but it is more than Conroe Duo. But to make a very quick summary of the article, it is this: cost.
The cost of replacing a 939 system with AM2 doesn't justify the price point.
Windows has detected an undetectable error.
AM2 really is an excellent platform, it consolidated AMD's Value, Mid-Range, and High-End market segments into a single platform. The reason it's not viable in the larger market-wide Enthusiast, Performance, and High-End segments is simply that Core 2 Duo rapes it. If you're already considering spending the money for a higher-end Athlon 64 X2 or FX processor, you can move to a Core 2 Duo-based platform that will destroy the AMD options performance-wise by a margin that is nearly unprecedented while still providing good power and heat usage. Basically, if the market was perfectly rational and had no transition times, all systems would be AMD AM2-based until you reached high enough prices that it was cost-effective to use a Core 2 Duo, and the P4 and Celerons would be merely a bad memory. AMD's aquisition of ATI helps it in this regard, as ATI has been making some chipsets that are very reliable, very fast, and rather inexpensive. ATI definitely has the best integrated graphics solution in the laptop market, and AMD's Turion 64 X2 is more competitive here than the Athlon 64 X2 is in the desktop arena.
Conroe (VHS) gives you more for less than AMD(Beta)'s superior Hypertransport and on-cpu memory controller. Conroe entirely stole the thunder of AM2, and consequently AM3.
When you can get a Core 2 Duo E6600 and have it crush an FX-62 and at a fraction of a FX-62's price... It's the same formula as always, price to bang. You get more bang with 939, or go straight to Core 2 Duo.
You could always argue time. AMD folks are used to living a long time on a socket type. 939 was only around about a year before AM2 came, whereas 754 and the previous socket 7 were very, very long lived. In another couple years, maybe AM2/3 will pick up steam, but it's too early.
1) While virtualization is immensely useful to a small number of people, it is virtually useless to most end-users.
2) While DDR2 offers greatly increased bandwidth, it does so at the expense of latency, and in many common applications, doesn't really perform much (if any) better than the 128-bit DDR memory of the socket 939 Opterons did.
When you look at it that way, other than being more "future-compatible", there aren't really any benefits to *most* end users, and if there aren't any benefits, why would they upgrade?
The Athlon64/Opteron chips were popular because they were innovative in useful ways, which gave the end user something more for his money. The AM2 hasn't kept with that tradition.
steve
Oh, you're not stuck, you're just unable to let go of the onion rings.
Well, the Woodcrest is spanking the opterons at present, I just got done benchmarking the new AMD chip vs the Intel woodcrest for PRMAN and Shake rendering and the AM2 is between 20% slower and 50% slower for what we want to do and the Opteron 280's are about as fast (core for core) as our aging 2.8GHz Xeons
To add to that, our reliability record for AMD systems is mindblowingly shocking. Having purchased 65 Dual 280 Opterons, we've had problems with ~60% of them.
Indeed. I really dont get the point. AM2 is simply a platform change; basically just a couple of lines drawn differently on the motherboard. And a new memory standard that's just not that big a deal (and, iirc, the reason it was a big deal at the Socket A introduction was that ordinary SDRAM performance really sucked and nobody wanted to touch RDRAM with a ten foot pole even if they could afford it, creating a huge up market demand for that specific change).
I am not sure if the memory standard isn't a big deal. It probably helped Dell adopt AMD, since they need same memory (DDR2) for Intel boxes, so Dell won't have to have 2 suppliers for memory.
This new memory might help also with quad cores and beyond. Right now the single/dual core AM2 is not bandwith starved (tests give DDR2 an edge of 3-5%), but that might change with quad cores and beyond where HT and faster memory could supply the cores where Intel CPUs might starve with a shared bandwith of 1033 or 1333 MHZ.
Customers and motherboard vendors alike are simply annoyed by the permanent socket changes. Sockets are hardware APIs which these days shouldn't change for a decade and not within a year or so. Besides the performance increase from 939 to AM2 is so insignificant there's no reason to switch.
IMO the best what AMD could do is scrap AM2 and replace it with a socket which is able to plug in 939 (DDR) processors and possible DDR2/DDR4/DDRx processors. Since this will take some time AMD should release any AM2 processor parallel as 939 processors, else AMD will possibly loose some market share.
O. Wyss
See http://wyoguide.sf.net/papers/Cross-platform.html
This very thing is coming back to bite intel on the ass lately. The idiots who cant let go of the clockspeed thing. I had a notebook I was recently trying to sell, a guy that was interested was completely irate because I had the nerve to sell a 1.6ghz notebook for only a few dollars less than some other guy was selling a 2.0ghz system. He refused to comprehend that the one I was selling was a centrino while the other one he was looking at was an early p4 mobile and that mine was actually faster and cheaper. I finally grew so frustrated that I lied and told him I sold it already and reposted the ad with a new picture.