Enough of This Feud...
There were not so many major events in the x86 CPU sector this year: Intel released Pentium 4 on Prescott core and Pentium M on Dothan core (90 nm process technology) as well as announced the new Socket 775 platform. In its turn, AMD announced the Socket 939 platform and started the serial production of Athlon 64 (the same 90 nm process technology) in the last quarter. Thus, the two main leaders officially made a draw. In future chronicles 2004 will be described as "a year when both Intel and AMD offered serial processors, manufactured by the 90 nm process technology, and updated their platform list". However, everything looks different from the inside, because of the considerable differences in the levels, on which the events took place in these two competing companies.
"The main alternative" x86 CPU manufacturer has been reaping laurels this year: to all appearances, it managed to bed in and optimize the manufacturing technology for K8 chips, due to which there appeared more of them on the market. Frequency of the top model was successfully raised to 2.6 GHz (Athlon 64 FX-55). AMD64 command set support is still not very good, though: Windows had never been released for this platform, Microsoft is still feeding users with promises. But it's no secret that this OS is installed on the overwhelming majority of personal computers. However AMD processors are doing quite well even without this seemingly major trunk card. Future is future, but from the point of view of a common user, Athlon 64 currently remains a fast and [relatively] inexpensive 32-bit processor with "virtually supported" 64-bit operating mode. It's only virtually, because, as we have already mentioned, for most users it operates in 32-bit mode de facto. As we have already mentioned before in our reviews, top AMD processors are no inferior to their main Intel competitors in performance, and they even outscore them in some applications. Thus, having rid of the catching up syndrome, the company managed to pay more attention to the solution of other problems. What problems did AMD choose to solve?
Firstly, the new Socket 939 platform was announced. This change actually fits well the definition of "cosmetic", because you will need a high-powered microscope in order to distinguish between Socket 939 and Socket 940 on the desktop market. Even under the threat of being stoned for vexed questions I will say the following: Socket 939 is just what Socket 754 should have been like. The parallel with Socket 423 –> Socket 478 transition inevitably comes to mind, and with all divisiveness of this comparison... well, one will automatically make this comparison with all its disputable nature :). One can take one's time and evaluate pros and cons of Socket 754 compared to Socket 423 (the first platform for Pentium 4), but even the most ardent AMD fans agree that the 939 is much more promising than the 754, and it will most likely survive for a longer time. However, if users condoned Intel with its Socket 423 – they will have to condone AMD with its 754. Especially as all the grudges are in the past, while all the joys lie in present or in future: Socket 939 looks much better than its predecessor as a platform for a desktop Athlon 64. So, AMD is again at advantage here: CPU frequencies grow, the new desktop platform is a success.
The second large-scale innovation of the expiring year is the appearance of Sempron CPUs, AMD has obviously followed the Intel example here as well. The know-how consists in the reservation of CPU name even when the platform is changed. Intel did the trick with its Celeron brand: Cores and sockets would change – but the brand (processor name) remained the same. The first Celeron processors were released for Slot 1, then this name migrated to low-end CPUs for Socket 370, then – to Socket 478, and now (with D added at the end) Celeron successfully copes with Socket 775. Like a Volkswagen "Golf" – more of a symbol than a certain model. It's actually a good move, it takes into account the psychological differences between potential buyers of middle/high-end and low-end solutions. The latter usually cares not about the upgradability and lifespan of the system – he needs the best performance for his money right now. And so the difference between the platforms is considered mostly in terms of money: if not pressed for money, one can choose Sempron + Socket 754 with some extra performance in reserve, if on the rocks – one can buy Sempron + Socket A. At least you can boast of Sempron to your friends and relatives, being too modest to elaborate for which socket it is :). On the whole, it's a good idea, if not pay attention to the traditional disappointments of some fans, who considered themselves "fooled" by Sempron indices. However, these indices are also quite clear, if your use your common sense: if this processor brand is considered by AMD as the alternative to Intel Celeron – it's quite logical that its indices are "tied" to the main competitor instead of Pentium 4. In short, it's another advantage from the point of view of a common user, not a fan who is expecting a moon for a sixpence from the favourite company.
The 90-nm process technology from AMD is slightly out of focus in analyses and forecasts. Judging from some publications, the 90-nm process technology offers considerably reduced heat dissipation, which is a very nice thing. However, the fact that top processors are still manufactured solely on the old technology is suspicious – it suggests an idea that the 90-nm process is still far from being fine-tuned as the 130-nm process. So let's tick against this item (processors are in serial production), and put an interrogatory next to it (let's wait until the 90-nm process is used in Athlon 64 FX, for example). One the whole (which is well clear from the above said) 2004 was a successful year for AMD, though the company made several steps, which changed its image among users from "the best friend of the unfortunate and the oppressed" who tilts at windmills to a normal respectable manufacturer (even if not the largest on the market), who cares about its financial success and descends to professional marketing moves (even if they are partially adopted from the leader). I like the new AMD image better than the old one – it looks somewhat more reputable and safer. We all make money, let's not be ashamed of it...
We just cannot call this year successful or unsuccessful for this manufacturer. Turning-point, innovative, intricate – these are the words, but successful or unsuccessful are definitely not the terms for this situation. On the one hand, it must be confessed, there were a lot of disappointments, and the greatest one is Prescott core. It turned out generally no faster than Northwood at the same clock frequency, besides it is noticeably more prone to overheating (which is not denied by Intel). Besides, the disappointment in Prescott indirectly cast an imputation on the character of the 90-nm process technology from Intel, at least before the release of Pentium M on Dothan core. Fortunately, the latter confirmed that the process is OK, it's just that the Prescott core is so hot in itself, not concerning the process technology. Whatever the acquittal, the aftertaste remained.
The new LGA775 platform also excites mixed feelings, and you should have a clear idea of the reasons: that's because it is new. The fact is: users do not accept the laws of logic. I want DDR2, but so that DDR remains! I want PCI Express, but leave PCI and AGP in place! No matter if you cannot imagine such a motherboard, because it wouldn't have free space for a processor: I told you – users do not use logic, they only have a categorical imperative: "I want it!" And though Intel acted absolutely right (I categorically insist on it!) – the new Socket 775 platform reaped full-scale its share of indignation and protests. But on the other hand, Intel certainly knew what it was about to do and did it deliberately. And again I take the floor: it did the right thing. In return we have a platform, which main components (DDR2, PCI Express, Serial ATA, USB, LGA775 socket, 24/8-pin power connector) will live rather long, and hopefully without major modifications. However, this set lacks new PC cases of the BTX format, but it was probably decided not to overdo with the number of innovations at a time.
And finally the notorious Processor Numbers. Firstly, Intel lost its face only because PN was taken as the counterpart of AMD ratings. I can say this: users will not look into any issues. That's why "if PN looks like an AMD rating – then it's a rating". Unbreakable logic ('cause there is nothing to break...) Secondly, PN was introduced at the same time the Socket 775 was launched, and the main processor for this platform is Pentium 4 on Prescott core. It goes without saying that the witch hunters "smelled a rat" right away: PN is introduced for LGA775 processors to do a number on users' heads and hide their low productivity. They traditionally ignored the explanations that not only Prescott is outscored by Northwood on some tasks, but the contrary situation also takes place; that these cores generally offer the same performance; that Northwood has reached its maximum frequency, while it will still grow in Prescott; that Prescott is cheaper (for customers as well, especially LGA775 modifications); and finally, that the frequency will be written on the box anyway, just in smaller font size. That's because the explanations are very lengthy! Yep! While "Intel is trying to cheat us!!!" is much shorter a phrase. It offers more zeal and is easier to remember. And of course, these three exclamatory marks at the end are mandatory: you cannot contravence traditions...
Though it should be noted that even Intel "loves" its new platform in a very strange way: Not only it initially provided no 1066 MHz FSB support, the first processor for this platform was strange as well: Pentium 4 Extreme Edition on the old Gallatin core (the Northwood with L3 Cache), 3.4 GHz only. That is, to put it simply: the first processor for the 1066 MHz bus needed such a bus least of all. However, no doubts, in 2005 there will appear Prescott with the 1066 MHz FSB. And, by the way, remember this assumption – you will need it at the end of this section. Approximately the same situation is with EM64T support (Intel's counterpart of AMD64) – processors seem to have appeared, but only for servers. One more time: it's clear that servers and powerful workstations are the ones that need 64 bit addressing more than any other computer types. It's clear to me, may be some readers understand it as well... But that will not allow you to buy a home desktop computer with 64 bit processor from Intel and boast of it to all your friends and acquaintances! While AMD offers not only Opteron but Athlon 64 as well...
Intel had no luck this year. Each bottle of ointment had its fly, and clean bottles went unnoticed. Let's take for example the continuation of Pentium M series – Dothan core. Good processor? Yep, just excellent! And when do you remember it among other things concerning Intel in a conversation? Right, because in this case Intel outscored just itself. It had a good Banias core – now it has a good Dothan core. It's not interesting – no plot, little passion ... But let's be frank, the global fault in the Intel-client relationships this year is Prescott. After its grand disillusionment all other Intel initiatives were given hostile reception. The funniest thing is that Prescott is not that bad. Applications with Hyper-Threading support demonstrate high gains on this core. The frequency really grows: it has reached 3.8 GHz already, while Northwood is frozen at 3.4. But higher heating is compensated: the processor starts throttling at a tad higher temperature and it does it better at that. SSE3 command set is quite a good supplement to SSE2 and is used in many programs already. According to low-level tests (RMMA), the memory management is also better implemented – improved Prefetch algorithms and BIU unit in particular. Thus, Prescott is disliked because it is worse than expected but not because it is that bad. That's also very "logical", by the way...
However, despite all above roughness, Intel managed to achieve much in the expiring year. Firstly, to launch the serial production of the new core. It should be crystal clear here that Prescott is not a counterpart of 90 nm AMD Athlon 64. The 90 nm K8 core is practically an exact copy of the 130 nm one. The only essential difference is actually the process technology. But Prescott differs much from Northwood. In fact it's more like "Pentium 4.5" than Pentium 4. Secondly, there appeared a new platform with all stakes on PCI Express and DDR2. Thirdly, there was introduced Processor Number, which was really necessary despite the predictable negative reaction of some consumers. We've got an interesting situation here: it looks like Intel deliberately sacrificed 2004 for all its ambiguous steps to be able to start 2005 dressed in white. Something like: "Let bygones be bygones, it's January 1, 2005. Look at this: what an excellent series we have! What a nice platform! What progressive technologies! All of it is offered only by our company!". And here we go – "tasty things" in a long train: new Prescott for 1066 MHz FSB, new frequencies (you never can tell – the milestone 4 GHz processor may be back in the roadmap...), desktop processors supporting EM64T, and so long and so forth. If that was really the plan – I bow my head: it's a good plan. It really explains almost everything. What concerns a year of "ill luck"... Well, what's a year for such an influential company? It will overcome. In fact, it has already overcome it.
Industry on the whole
If you are thirty, you wake up, and nothing hurts...
...Then you are dead. It's an old joke, but it's quite appropriate here: the processor industry is not that young. And one of the odious delusions, constantly entrapping some users (or even journalists – for a melodramatic statement and cheap popularity), is the opinion that "everything is bad these days, no one is trying to solve users' problems". While in fact the truth lies on the surface: the older the industry is – the heavier is the load of problems, which are hard to solve or cannot be solved at all at the current stage of development. One can lament over high power consumption and "nightmarish temperatures" of the present CPU, but no one will abolish the law – "the more transistors and the higher frequency, the higher the temperature". Some users manage to solve this problem in a tad better way, the others – a tad less effectively, but sooner or later the power consumption in the next model will grow anyway, as well as its heating. And again: one can argue much about how faster DDR2-533 is than DDR400. However I hope no one doubts that both memory types are much faster than DDR266 and that really has its effect on application speed. Look back more often: when you get used to speed, it doesn't seem high any more. But looking back, you can get an idea how far we really are from the starting point.
I just love one simple example, which I always tell to those who like to remember "the good old times": if we take the Pentium II series for example (all the cores inclusive, we shall not peddle the details), then the frequency of the top representative of this series related to the junior model frequency as 450:233=1.9 that is the senior model outscored the junior model only by 1.9 times. For Pentium III this ratio will be 3.1 (1400:450). The current Pentium 4 reached the 3800 MHz frequency with their minimum of 1300 MHz, that is the scalability ratio of the base core architecture is now equal to 2.9. Then what are you crying at, may I ask? Were those times that "good"? Well, it's always good for you to look back...
But now practically no one doubts that the frequencies will grow more slowly. Well, they'll try to find other ways to raise the performance. Frequency is not the only pebble on the beach after all. What's important – we should awake to one simple fact: wailing and crying, or applauding and rejoicing, or in a usual humdrum way – the industry progresses anyway. And all these wails and applause actually have no considerable effect on the progress rates. It's naive to think that any "expression of mass indignation" over delaying releases of new CPUs will accelerate their development or increase the factory output: people clamour, clamoured, and will clamour – but engineers work like they always did. Most likely this entire clamour even doesn't reach their ears :).
Alas, the first sign of "mature" industry is a scheduled release of products, which is profitable to this industry. CPU industry is now witnessing pragmatic managers and marketing specialists succeeding to "the gifted and geniuses". When this process is completed, we most likely won't hear any more about "technological break-throughs" loved so much by IT-fans and mass media (in my case "breakthrough" always associates with sewerage...), but we'll gain the confidence instead that tomorrow will always be slightly better than yesterday. Sometimes at the cost of artificial slow down of the progress, but that's the price of stability: out of three discoveries made simultaneously, only one is immediately applied in industry... because the next two years may have no discoveries at all, but "one's to make a living" :). There is nothing terrible about that: we are long accustomed to the same behaviour in the automobile industry, consumer electronics market, and even clothes. Accustomed, and even don't notice it. That's because in fact we are quite satisfied with the quality and functionality of today's cars, jeans, fridges, and TV-sets. That's actually the idea: being used to pragmatic attitude to these things, we are not waiting for them to work magic, just normal operations. Only toys can be willingly replaced each day, while old comfortable shoes are much better. At least until worn into holes...
Right of choice, or
|blog comments powered by Disqus|
|Most Popular Reviews||More RSS|
Comparing old, cheap solutions from AMD with new, budget offerings from Intel.
February 1, 2013 · Processor Roundups
A couple of mid-range adapters with original cooling systems.
January 30, 2013 · Video cards: NVIDIA GPUs
An external X-Fi solution in tests.
September 9, 2008 · Sound Cards
The first worthwhile Piledriver CPU.
September 11, 2012 · Processors: AMD
Trying out the new method.
September 18, 2012 · Processors: Intel
|Latest Reviews||More RSS|
Retested all graphics cards with the new drivers.
Oct 18, 2013 · 3Digests
Added new benchmarks: BioShock Infinite and Metro: Last Light.
Sep 06, 2013 · 3Digests
Added the test results of NVIDIA GeForce GTX 760 and AMD Radeon HD 7730.
Aug 05, 2013 · 3Digests
An excellent hybrid of GeForce GTX 650 Ti and GeForce GTX 660.
Jun 24, 2013 · Video cards: NVIDIA GPUs
Added the test results of NVIDIA GeForce GTX 770/780.
Jun 03, 2013 · 3Digests
|Latest News||More RSS|