Original Link: https://www.anandtech.com/show/1213
Price Guides December 2003: The Year in Component Prices
by Kristopher Kubicki on December 25, 2003 11:02 PM EST- Posted in
- Guides
It was about this time last year the Pentium 4 2.4GHz 533FSB chip became the Intel chip to own. Funny enough, the 2.4GHz 533FSB and 800FSB CPUs continued to dominate the sub $200 price range for the majority of 2003. Only recently did the 2.6GHz 800FSB chips fall into the ~$170 market.
It is very easy to imagine the 2.8C will fall into that sub $200 category very soon; our price analysis has shown the processor prices are still on the move since the October 24th Intel price cuts. It feels almost a lifetime ago that the 533FSB 2.8B was $450 processor (December 2002). However, the most obvious trend here is the lack of ramping for the Pentium 4 line. The Celeron line ramped quite a bit, but we all know that decision on behalf of Intel was quite pointless. In fact, the only real addition to the Pentium 4 line was HyperThreading, 800FSB and the 3.2GHz P4s.
HyperThreading was an excellent addition to the P4 line, although sometimes it can be more of a hindrance in certain situations. 800FSB was more of an evolutionary advancement rather than revolutionary however. If we want to talk about revolutionary, we only need to go around the corner to our friends in Sunnyvale.
With Prescott around the corner, it would be interesting to see what Intel has been doing for a little over a year.
AMD was quite the busy bee this year. Twelve months ago, 333FSB was brand new to the 2600+ and 2700+ processors, and KT400/nForce2 were the only fast FSB options. Since then the introduction of the Barton core (particularly the 2500+) set the bar for price and performance in the CPU industry. In the last four months, the Barton 2500+ has continually proven itself as the best processor available for under $100 (and even arguably one of the better processors for under $200).
A year ago, the Athlon MP line was just beginning to ramp up into the 2200+, 2400+ speeds; but it was probably too little too late. We only need to look at how quickly the Opteron line replaced Athlon MP to judge how badly the MP line needed a replacement. Along those lines; 2003 was the year of the Opteron. Once we saw the introduction of the Opteron around May/June, the chip became the paragon of price, performance and reliability. Recently, the chip ramped up to its fourth speed revision (X48) and doesn't show any signs of slowing down.
Unofficial rumors of FX-53 and FX-55, as well as half a dozen ramps to the Athlon 64 3000+ to 4200+ processors look extremely convincing. The San Diego line looks to become the most promising Athlon yet, and we can't wait.
Earlier this year, AMD made some attempts to penetrate Intel's grip on the high end market with its Athlon 3200+ and 3000+ 400FSB CPUs, but alas again it was too little too late. As if almost an exact repeat of the Opteron launch, the Athlon 64 line quickly gained recognition as the best available CPU bar none. Of course, that was further compounded by AMD's silent release of the Athlon 64 3000+ Newcastle processors also known as "the best processors AMD never announced."
In fact, the Socket 754 Athlon 3000+ is our recommendation for new processors this week. At $240 it costs less than Intel's 3.0C, but it also thoroughly outperforms the Intel chip (even without 64bit optimizations!). Considering the best high end chips of last year cost $500+, and still could not outperform the "low end" Athlon 64 3000+ of today; it appears the CPU market is making excellent strides in price and performance.
In December 2002, there were not very many choices as far as Intel motherboards go. SiS and Intel were both battling it out over the 533FSB market with i845PE and SiS 648 although Intel was clearly coming out ahead. Intel's extremely short lived Granite Bay chipset came and went with a 5 month product cycle; certainly a trend we would not like to see occur too often. Many of you might also remember Intel's sneaky move where they put all 865PE advertisements and announcements under non-disclosure agreements even though the boards sat on store shelves. As a result, the near identical performance of 865PE and 875P boards succumbed to clever marketing and 875P boards flew off the shelves.
Seven months later, not nearly as many motherboards companies have unlocked 865PE chips left and there actually is a performance difference between 865PE and 875P. The bigger Tier 2 companies (Abit, DFI, Albatron) should still have enough unlocked 865PE chips to continue their board production, but that has yet to be confirmed.
SiS and VIA are still both in the Intel chipset making business; SiS's fairly robust 655TX boards are supposed to be some of the best available. However, as always seems to be the case with SiS's Intel struggle, it may be too little too late. VIA has been slowly growing recognition for its PT800/PT880 core, although many people have been reluctant to adopt the alternative chipsets over 865/875. Intel is just holding too many good cards in its hand. Perhaps VIA's partnership with S3 will secure a stronger position in the Intel IGP field.
Speaking of Intel IGP, the little engine that could (ATI), made quiet incredible headway this year with its 9100 IGP RS300. Of course its performance could not match that of 875P or even 865PE, but the integrated Radeon 9100 does make short work of any integrated graphics from VIA or Intel. We will be seeing a lot more of ATI in the core logic sector of the future.
AMD motherboards underwent few changes in the Socket A department over the last 12 months. The largest change came in the form of 400FSB chipsets, both from NVIDIA and VIA. Of course, most of us already knew that the original nForce2 chipsets were in fact 400FSB capable, designed from the beginning for expansion. VIA, on the other hand, had some severe trouble getting up to 400FSB. KT333, KT400, KT400A and then finally KT600 all came within 12 months of each other.
But then, the quintessential problem plagued both avenues of 400FSB; what is the purpose of 400FSB if there are no 400FSB chips? Athlon XP 3200+ became the first 400FSB AMD processor and although it was welcomed, the cost was prohibitive. A 3000+ 400FSB processor was released without any PR endorsements (think Athlon 64 3000+), but prices on that processor were also prohibitive. Even today, the cost of a 3000+ 400FSB XP processor costs the same as a Pentium4 2.8C.
All in all, the Athlon XP market for 2003 was a mess. Barton processors were the year's only saving grace. In the end, nForce2 became the better overclocker chipset over VIA's KT-anything. nForce2 has undergone some minor tweaks; particularly there have been a few (annoying) stepping changes in the last two months. However, like with the difference between steppings in the 865PE chips, the bigger Tier 2 motherboard manufacturers have enough originally stepped nForce2/Ultra/400 chips around to assure solid performance (Abit, Shuttle, Soltek).
One of the things that hurt VIA the most was nForce2's "competitive" dual channel memory strategy. Although there were some cases where dual channel memory provided a performance increase over single channel, around February/March of 2003 people began to start realizing single channel controllers could in fact be just as competitive. After VIA lost that marketing war, the nForce2 boards had matured enough with onboard gigabit, digital coax/optical audio, and most importantly, integrated graphics. VIA recently announced its dual channel Socket A core logic, but with the advent of extremely cheap Athlon 64's, perhaps VIA is better off dropping its pursuit into the Athlon XP market and continuing doing what it is already doing the best; Athlon 64.
Even though competition between nForce3 and K8T800 has been pretty neck and neck so far, VIA seems slightly ahead of NVIDIA as far as price. There has been no shortage of adoption of either chip, but as you can see VIA has just a few more motherboard supporters.
VIA has a few key advantages over NVIDIA. Primarily, CK8 (nForce3) has been sitting on the shelves for years. Even at COMDEX 2002 we were already getting a little tired of hearing about the chip. VIA started fresh and did not debut into the mainstream market until the Athlon 64 announcements in September. Second, the largest issue VIA had in the K7 architecture was the memory controller, which AMD moved onto the processor for the K8. This also attributes to SiS, ALi and ATI's announcements to move into the AMD core logic sector.
With NVIDIA and VIA both announcing second generation Athlon 64 core logics at CES 2004 (next week), it will be interesting to see how the market heats up. SiS and ALi are still working to mainstream their first generation chips, and ATI hasn't even made their product publicly available yet.
Memory went under some very important changes this year. Primarily, RDRAM disappeared with Intel's lack of commitment to keep the architecture alive. With DDR speeds increasing so rapidly and cheaply, Intel's decision to basically abandon anything Rambus was probably an excellent decision. The incredible success of 865PE and 875P became the final nail on the coffin for RDRAM.
From December of last year, memory speeds increased almost 100%. DDR400 just began to show up in the mainstream (although the format standard was not sanctioned). Mushkin and OCZ are both approaching the DDR600+ threshold, which has done incredible things to the memory market. As speeds continue to increase that high, slower speeds such as DDR400 are forced to lower in price to stay competitive. We only need to look at the last years prices to see this force in action cost approximately 3X today's prices (and with slower timings).
The disappearance of PC133 was a particularly interesting trend. At a memory conference in March, we asked a high end memory manufacturer about the consequences of DDR effectively replacing PC133 in less than 6 months. Unfortunately, if history is any lesson, a similar trend will probably occur within the next 18 months as we start to see the introduction of DDR2. We will just have to see what the next year brings for us!
Hard drives are typically the most conservative of the major computer component sectors. We did not take an active interest in tracking hard drives prices until three months ago, but that does not mean we don't know what happened with hard drives this year.
SATA has been the continuing delta in the world of magnetic storage. A year ago SATA drives held a $100 - $200 premium over any IDE counterpart. Today, there is less than a 10% premium on these drives; we would not be surprised if the drives become more prolific than IDE by the end of 2004.
SATA continues to become the better choice over IDE. This year we will definitely see an increase on the bus speed. Most motherboards since May of 2003 began shipping with SATA connectors.
Unfortunately, there has also been a darker side of the magnetic storage sector. Continual reductions in warrantees, quality and longevity have angered consumers and retailers alike. One needs only to browse our forums for a taste of such unfortunate conditions.
For video cards, the last 12 months have been a battle zone. ATI did some serious damage to NVIDIA with the Radeon line in late 2002. NVIDIA's response with GeForceFX was perhaps not their best effort. As we all know, the GeForceFX 5800 was almost immediately replaced with the easier to product 5900 line. Unfortunately, less than 6 months later that line was replaced with the 5700 and 5950 lines. If this was core logic instead of GPUs, we could have easily mistaken NVIDIA for VIA.
What makes matters more confusing and perhaps less credible on behalf of NVIDIA was the introduction of various notations on their cards. The "Ultra" cards directly correlate with ATI's "Pro" notation; however other additions like XT and SE do not. The GeForceFX 5600 XT has no competition with a Radeon 9600 XT, and we feel the addition of such modifiers has basically only been to confuse consumers. NVIDIA has some excellent video cards, but unfortunately they are clearly banking on consumer ignorance and their own name recognition over ATI.
There are still some excellent NVIDIA cards. Our particular favorite, the GeForceFX 5900 NON-Ultra has been a spectacular sub-$200 card for several months now.
Then there is ATI. ATI isn't exactly free of guilt when it comes to marketing and consumer confusion, but they are considerably more straightforward than NVIDIA. Unfortunately, the year 2003 brought the demise of one of our favorite video cards, the 9500 Pro. These were great cards that could easily overclock and achieve near 9700 Pro capabilities for half the cost.
Furthermore, with the introduction of the Radeon 9800 Pro, and then 9800 XT, ATI began to phase out its 9700 Pro line. For around $220 the 9700 Pro is still the best card money can buy and it certainly gives any $300+ card a run for its money. Sources continue to dry up, and it's only easy to say by Q2 2004 this card will go the way of the 9500 Pro.
ATI made much headway in the mid and low range market as well. We were not pleased that the 9600 Pro eventually became slower than the 9500 Pro (consumer confusion), but we were relatively happy with the performance. Of course the lack of any true DirectX 9 games have been extremely discouraging. Half Life 2, the DirectX 9 spearhead, had an unfortunate problem that ended up delaying its release further than it already was. As ATI and NVIDIA (NV40) gear up to tape out another chipset and take yet another swing at each other, consumers watch idle; confused.
We hope you enjoyed our brief analysis of the year in component pricing. From the incredible success of the Opteron launch to the dismal defeat of NV30, 2003 produced some of the best and worst investments of the computer industry. It is clear that the volatility of this industry obtained a level we have not seen in the past. Was 2003 an omen to future component prices, or an unfortunate fluke that will subside in 2004?
Facebook
Twitter
RSS
