Processor results of 2008

Processors and chipsets

Intel confidently entered 2008, having in its portfolio not only quad-core Core 2 Extreme QX6850 processors boosted to 3 GHz, but also their successors, the first of which (Core 2 Extreme QX9650 , also at 3 GHz, but with a more capacious second-level cache) went on sale back in November 2007. As usual, Intel’s top models, with the word “Extreme” in their names, cost a lot of money (up to $1,000 in wholesale quantities), but their appearance always means only one thing: the technology has been worked out, the factories have been re-equipped, reconfigured and restarted, you have to wait in the coming months the appearance of mid-price stones from the same line. And remembering that any of the “Extreme” Intel processors received a mid-end analogue no more than six months later, inferior to the extreme version only in the absence of an unlocked multiplier, all fans of Core processors were optimistically awaiting the onset of processor communism.

Core 2 Quad in various incarnations is the hero of 2008. In 2009, it is replaced by the Core i7 family on the Olympus of productivity

Actually, until the fall it was more or less like this – all lines, classes and groups of processors for various platforms were gradually updated, and manufacturers of motherboards and chipsets, meanwhile, continued to scatter away from the yoke of Intel itself, which actually occupied market of chipsets for desktops and laptops of traditional format. Where should we run? That’s right, into new, not yet fully developed market niches. As a rule, they are miniature… However, we’ll talk about platforms a little later.

 

Phenom X3 and Phenom X4 did not bring AMD into the lead, but forced Intel to make room in the mid-market segment

What about AMD? At the beginning of 2008, no one could say anything definite. The merged company was having a hard time on almost all fronts; competitors were also pushing it out in the video chip market, despite the fact that even when purchasing the Canadian company ATI, the development of the latter’s new ultra-high-performance R600 chip was supposedly almost completed. However, we all remember how long we had to wait before the R600 finally appeared on the market in the form of the Radeon HD 2900 XT. We also remember the shortcomings of this video card, which were more or less corrected in the Radeon HD 3000 series, which was also a good alternative to nVidia products, but nothing more. Serious competition in the top segment at the beginning of 2008 could only be discussed with reservations about CrossFireX modes or multi-chip video cards.

In general, in order not to go too far aside, let’s just summarize that AMD entered 2008 with serious problems on all fronts and the company clearly did not have enough resources to solve all the problems in parallel and in the shortest possible time. However, the long-awaited breakthrough with the Radeon HD 4000 took place in the best possible way, and the second generation of Phenom, created on a 45-nm process technology with the help of IBM engineers, has already shown its potential at closed presentations. We are looking forward to February, when in our test lab we will see not engineering samples of new stones from AMD, but production models available to “mere mortals”. It is unlikely that Phenom II will throw off the throne the newly baked king named Core i7, but no one is stopping you from filing a couple of legs, so that later, when the opportunity presents itself, you can still do your dirty deed.

AMD’s plan shown two years ago. Now we are between the second and third stages

There was also plenty of news regarding chipsets and motherboards based on these chipsets, and not all of them were strictly tied to the release of new generations of processors. This market was developing in all directions, and in connection with the emergence of a new class of computers (and more than one!) I had to remember the seemingly forgotten names VIA and S3. Although, again, there is nothing to be surprised about – after the ousting of “large” PCs from the market, their domain became mobile and ultra-compact solutions, which began to fill the shelves like an avalanche in 2008.

From global to local

Somehow, thought constantly runs away from a consistent analysis of the events of 2008 to the revolutions that took place at its end. Of course, there is nothing to be surprised here, the memory of recent events is still fresh, and what happened a year ago seems incredibly distant – IT life is generally a rather dynamic thing. And yet, it is still necessary to trace the main events in chronological order.

The beginning of 2008 in the processor market was an important turning point in the fate of quad-core processors. Of course, the Core 2 Extreme QX6700 has been in production since November 2006, and its “affordable” brother, the Core 2 Quad Q6600, appeared in January 2007. But it was too early to talk seriously about the victorious march of quad-core processors. After all, the Q6600’s $800 starting price didn’t do much to help its popularity.

The situation could only be changed by a worthy competitive offer at a lower price, which is what the Phenoms released at the end of 2007 became. At that time, AMD processor manufacturing technology did not allow for a “gigahertz war,” but price was always welcome. In this matter, AMD has always performed well, for which even the most ardent Intel fans owe it a huge thank you. If it were not for the release of Phenoms at the end of 2007, then, most likely, for the entire 2008, prices for “quads” would not have fallen below $300-400. And so – back in January, the Phenom X4 9500 was sold for $200 and then, naturally, it only became cheaper. The rest of the quad-cores automatically, one after another, also began to “move” below this psychological boundary.

What is the price? Less than $200? Four cores? Damn tempting.

Now, de facto, quad-core processors have become a mainstream product, the price of which is much lower than two dual-core processors of the same frequency. At the moment, not everyone will be able to take advantage of this ratio – too many software (and, first of all, games) remain at most “dual-core”, but the price already allows you to take four cores “in reserve”. And there’s nothing to say about those who process 3D and real-time media streams. For them, 4 cores are a pure profit.

By the way, budget profit for the same users working with software optimized for multi-threading, in the summer there was another reason for joy – Phenom X3 processors. Individually, the cores of the new AMD architecture were inferior to the cores from Intel, but the pricing policy of the “green” ones brought 3 cores and 2 cores almost head-to-head. And in this situation, the benchmark results depended solely on their optimization. Result? It is obvious – everyone who needed good power for heavy programs with minimal financial outlay bought the X3, the rest were happy about the even greater reduction in prices for dual-core processors. Everyone is fine, everyone is happy. And with that, there was no special news from AMD until the fall – there was a systematic increase in frequencies and expansion of the model range, everything was as usual.

Yes, I completely forgot to mention the “TLB problem” that was sensational at the beginning of the year, which could cause a critical error on systems with the first Phenom models. The news then read like reports from the front, BIOS updates, designed to programmatically bypass this AMD error, reduced overall performance by 10-15%… Of course, this whole story (or rather, its exaggeration in the mass media) spoiled the initial sales of Phenom’ ov, however, from personal experience of using a processor with a TLB error in our laboratory (for two weeks), I can only say that no manifestations of this flaw were noticed. Even then, we wrote that when a new microarchitecture is released, there are always enough such errors. In the same first generation of Intel Core there were errors like fleas on a dog, and there was noise in the media. Just a little quieter. And the quick release of Core 2 completely stopped all the gossip.

AMD, of course, could not release the second generation of Phenoms immediately after the first, but the decision to add “50” to the indexes of all corrected processors (released with the new stepping) made it possible to avoid confusion and reassure the majority of doubters. As a result, if you decide to buy a Phenom X4 now, then an index like 9850 or 9650 (the main thing is that it ends in “50”) completely guarantees you the absence of a critical error in its architecture. At the same time, if you suddenly come across one of the Phenoms with indexes divisible by 100 (for example, Phenom X4 9500), you should only be wary if the processor is planned to be used in an important server. In all other cases, the likelihood of freezes and crashes due to glitches in the operating system or the application itself will still be much higher.

What did Intel do at the beginning of the year? Yes, still the same – it produced expensive but incredibly powerful Extreme processors and updated its lineup, replacing old 65 nm processors with 45 nm ones.

In this segment, the Core 2 E6000, E4000 and E2000 were replaced by models from the E8000, E7000 and E5000 series. The frequencies have increased a little, power consumption has decreased, the L2 cache has become larger (6 MB on the E8000 versus 4 MB on the E6000), and so on. In theory, the 45-nm process technology makes it possible to obtain cheaper processors, and the prices announced by Intel were indeed low. However, a slight artificial maintenance of the shortage of 45-nm processors did its job and at retail the new Core 2 were noticeably more expensive. However, there is a logic to this and it is obvious: it was necessary to sell off the remaining 65-nanometer processors. For this reason, by the way, I very rarely advise buying products from market leaders immediately after their release. The lack of a worthy response from competitors relaxes not only the manufacturer itself, but also retailers. Yes, everyone wants to eat, but why should you and I pay for it? Let the impatient ones pay, and we can wait a month or two.

Compatibility with the old, well-deserved LGA775 socket made it possible to upgrade without any problems for all those who had taken care of a motherboard with a fast FSB in advance. But in general, the fact that the mainstream Core 2 Duo E8000 began to require a FSB frequency of 1333 MHz required the release of new chipsets – the 40th series. The 50th will appear only at the end of 2008, and will be designed to work with the new generation of Core i7 processors.

AMD does not have such problems. Actually, any of the chipsets that support Socket AM2+ are suitable for any of the Phenom processors and, most likely, will not have problems with the Phenom II. If only the power circuits on the motherboard support the necessary modes.

But AMD still has a serious gap in another direction – laptops. While Intel, as part of the promotion of the Centrino 2 platform, systematically updated Core 2 Duo T processors, replacing 7000 stones with 9000 stones, and also introduced the first mobile quad-core Core 2 QX 9300 and Core 2 Q9100 to the market, AMD remained in the shadows. If we take into account the delay between the development of desktop and mobile solutions from all manufacturers without exception, a fundamental change in the balance of power between laptop processors will not be expected soon. For now, AMD only has a more balanced platform – in terms of 3D processing speed… that is, games. Intel’s latest graphics chipsets (GMA X4000) have improved their performance with HD video, but these chips do not handle games much better than the previous 3000s. And those, in turn, are also not far from the 950. It’s obvious that Intel is targeting primarily serious people who would rather watch a movie for fun than waste 5-10 hours on a fresh shooter. On the other hand, casual games (The Sims 2, Spore, World of Warcraft and many others) run fine on Intel graphics and many are happy with this. But not everyone, of course, and AMD sees its chance in this. Yes, AMD mobile processors are weaker, but the graphics are a head (or even two) faster. And this is only the built-in one; we’ll talk about discrete laptop video chips separately.

Now let’s move on to what has excited not only the computer, but also related markets.

Peaceful Atom – for every home

There have been conversations about Intel’s mysterious project for ultra-compact PCs for a long time; in 2007, presentations regularly included information about how development was progressing, what had been achieved, and whether we should expect the release of a new processor on schedule. But interest in UMPC was frankly weak, and this is not surprising. Relatively expensive, not very functional devices, sometimes with QWERTY keyboards, but not so good that you can type in the usual way. This means again “two-finger” typing, which, in general, works fine on smartphones (unless, of course, your fingers are as thick as mine).

In 2008, the rapid growth of the netbook market noticeably slowed down sales of traditional laptops

In addition, early UMPCs had a criminally short battery life – about two hours. This, you know, is a joke for the chickens. Then this period was noticeably increased with the help of various tricks (each manufacturer in its own way), but the UMPC concept itself in the “screen with a sliding keyboard” format, without receiving the proper starting acceleration, stalled and settled in its narrow niche “for those who lack screen sizes of a smartphone or PDA.” Therefore, the new Atom processor, aimed at this specific niche, even despite its amazing energy efficiency indicators, was of little interest to anyone. How wrong we were all!

Under the beautiful code name Silverthorne was hidden a small (only 25 sq. mm versus 143 sq. mm for the Core 2 Duo) processor capable of performing only two operations per clock cycle. For comparison, let me remind you that the first generations of Pentium were distinguished by the same efficiency, and the huge performance gap between the Core microarchitecture and its predecessor Net-Burst (Pentium 4, Pentium D) was achieved not least due to the transition to performing 4 operations per clock instead of three. Thus, the Atom with a frequency of 1.6 GHz is inferior even to the Pentium 4 of the same frequency.

But honestly, tell me, what do you do on your computer that, say, a Pentium III with a frequency of 1.2 GHz could not handle? An office suite (without particularly heavy spell checking modules), a media player (though not for Full HD video), graphics packages (just so that the files are not too “heavy”)… Remember, six years ago we did all this on our computers and in didn’t blow a mustache. A modern machine can do everything two to three times faster, but here it’s worth remembering again the size of the Atom crystal. From a standard wafer with a diameter of 300 mm (45 nm process technology) you get about 2500 processors! This made it possible to start selling 230 Atoms for $29.

ВIntel Atom chip ID is not much like Core 2 Duo

AMD can respond to this with Sempron LE processors, which cost $30 and noticeably outperform Atoms in performance (by one and a half to two times). But the comparison of power consumption (45 W versus 2-4 W) leaves Sempron no chance in the mobile segment. Therefore, by the way, I don’t see much point in nettops (inexpensive desktop computers for web surfing) and home media servers based on Atom. Most of the time, their load on the processor is minimal, and in this mode the Sempron LE is only 3 watts behind the Atom. Another thing is netbooks, which are not faced with the task of “staying on 24 hours a day.” For them, it is more important to last as long as possible on one battery charge, performing a wide variety of tasks, often loading the system almost to its maximum. Atom is almost ideal for this.

Let’s not deny that Intel representatives immediately stated that they see in Atom a symmetrical response to AMD Geode, used in the first generation of OLPC – the notorious “100 bucks laptop for poor children from the third world.” But even at the time of the start of deliveries to OEM partners (in June 2008), few could have imagined what would come of this venture in the end. Let me remind you that the first Asus Eee turned out to be too expensive and inconvenient (primarily due to the miniature 7-inch screen), and AMD by that time had given up on OLPC and abandoned the production and development of Geode processors. It’s not for me to judge, I already mentioned above the problems facing AMD at the beginning of 2008, but now the “greens” will have to make up for lost time in the netbook market. And this is now a very “tasty” piece of the pie: if in 2007 only 1 million netbooks were sold in the world, then by the end of 2008 this number reached 14 million and continues to grow.

In the context of the global crisis, the advantages of netbooks suddenly became even more significant: low cost and efficiency with sufficient performance – many of those who were going to buy just a budget laptop will reconsider their intentions in favor of even cheaper options.

It is impossible not to mention another “half-hearted” solution from AMD – the Athlon 64 2000+ processor. From the name it is clearly clear that this is, in fact, a variation on the theme of the desktop Athlon 64 series, just with a lower frequency (in this case, up to 1 GHz). A full-fledged desktop architecture and the same full-fledged AMD 780G chipset allow the Athlon 64 2000+ to confidently compete with the Atom in terms of performance. Even the processor’s thermal package of 8 W looks quite tolerable, and a more energy-efficient chipset generally makes a system with an Athlon 64 2000+ more economical than a computer based on an Intel Atom. But the old 65nm process technology and large number of transistors mean that AMD’s ultra-efficient processor costs about $90, which is even three times more than the economical Semprons. If not for the price, the Athlon 64 2000+ would be the best solution for a nettop, but AMD, of course, will not produce processors at a loss.

On the other hand, the desktop dimensions (and the desktop AM2 socket) do not guarantee that this processor can be “stuck” into the miniature case of any netbook. It’s a pity, but nothing can be done about it. One good thing is that after the Phenom II hits the market, AMD will have every opportunity to make a more compelling response to Intel Atom.

And to close the topic of netbooks, let’s remember the third manufacturer, which has long been entrenched in the market of ultra-mobile PCs and simply compact systems – VIA. It was VIA that at one time introduced the first boards of the Micro-ATX, Mini-ATX and Nano-ATX form factors, and the VIA C7 and C3 processors for a long time had no equal in terms of efficiency.

With the release of Intel Atom, VIA unexpectedly abandoned the “arms race” in terms of efficiency and released the Nano processor, which received a larger cache volume (1 MB versus 512 KB) and a higher frequency (up to 1.8 GHz) compared to its previous solutions. As a result, the thermal package reached a value of 25 W, and this cannot be leveled out due to the chipset. A system with Nano under load consumes one and a half, or even twice as much as a system with Atom.

But Nano turns out to be higher in performance, even without taking into account the unique built-in random number generator and encryption unit (it’s not a fact that your software will support them). The situation leans towards Atom only in one case – when the program supports multithreading. Still, Intel’s decision to add support for Hyper Threading was correct.

Interestingly, in cryptographic tests the Nano has no equal at all, outperforming even the quad-core Core 2 (www.ixbt.com/news/all/index.shtml?11/17/22). But this is nothing more than another demonstration of the advantages of hardware implementations over software ones. It is important for a computer CPU to be universal, to be able to count anything. And you can’t get enough special blocks for all the tasks. In the history of computer technology, at one time only one task was separated from the central processor – processing three-dimensional graphics in real time (although even now, if necessary, any Core 2 or Phenom can easily draw graphics of three-dimensional games from the era “before DirectX, OpenGL and even glide”) . Now they are actively trying to “unscrew” physics, but judging by the modest successes of the Ageia company (the same one that was eventually bought up by nVidia), this process is still far from complete, and it is not a fact that discrete physics accelerators (even in the form of separate video cards in SLI or CrossFire combination) will become popular.

So, a cryptographic module is great, but neither office, nor Photoshop, nor WinRAR, nor media codecs will work faster because of its presence.

However, VIA Nano can still confidently compete with Intel Atom, but AMD has yet to come up with something new. Despite the fact that the processor codenamed Bobcat (which should become a direct competitor to Atom), mentioned in some news, is still being developed, according to AMD representatives, most likely, its capabilities will be implemented within the framework of Fusion (a universal chip that performs the functions of a CPU, GPU and the south bridge of the chipset at the same time. The reason for this suspicion is obvious – no matter how rapidly the netbook market (and, in addition, nettops) develops, in terms of mass it is still far from the traditional desktop/laptop sectors. AMD is unlikely to rush headlong into unexplored spaces, getting into trouble and spending huge amounts of money on PR and the fight against “childhood diseases” of the very concept of cheap and energy-efficient computers of small power (the first victim, by the way, was Asus, thanks to which we know, but a 7-inch screen is still too small for netbook). As in the case of DDR3, AMD can calmly wait and watch… And then Intel itself will move to single or dual-chip solutions (www.overclockers.com.ua/news/hardware/2009-01-09/103123) , which are more than justified in mobile devices. So I’m betting on Fusion.

Capons with artichokes and ginger

The past year will be remembered for several more exotic “iron” novelties, which remained exotic, not intended for the mass market. However, this does not make them any less interesting, since the solutions used to create such devices are often returned as architectural elements in serial products.

For example, the Intel Skulltrail platform presented at the beginning of the year, although it was declared as a “solution for enthusiasts,” in fact rather demonstrated Intel’s potential in the server segment. Of course, the presence of 4 PCI Express x16 slots made it possible to install almost any of the possible configurations of video accelerators (CrossFireX with four Radeon video cards, 3-way SLI with three single-chip GeForce or the version used in the system that Intel brought for demonstration in Kyiv – with two dual-chip GeForce 9800 GX2). But at the same time, the presence of two LGA775 connectors that support any Core 2 processors (including 4-core ones) looks redundant for games even now, a year later. So, the demonstration of the joint operation of two Core 2 Extreme QX9650 overclocked to 4 GHz was more likely to impress corporate clients who doubt the capabilities of the FSB bus to increase the processor power of the system, as well as professionals involved in video, audio and 3D processing.

However, to be honest, enthusiasts were also impressed, but there are not so many enthusiasts in the world who are ready to shell out $600 for just one motherboard (Intel D5400XS). Despite the fact that setting overclocking records on it is not as convenient as on single-processor systems.

Let’s also not forget that formally Skulltrail was a response to the AMD Quad FX platform, which, in fact, at one time (early 2007) was supposed to fill the gaps in the top segment simultaneously in two directions. At that time, AMD did not have either quad-core processors (Quad FX had two dual-core Athlon FX installed) or top-end video cards (they were supposed to be replaced by CrossFire combinations of up to four video cards). But, characteristically, such a monstrous design was relatively inexpensive – in the basic configuration, about $3,000. This is despite the fact that for several years in a row top-end processors from Intel cost $1,000 (well, okay, to be precise, then 999), and until recently the most powerful cards from nVidia were not sold much cheaper.

In general, both Quad FX and Skulltrail are interesting solutions, but at the moment they are too specific to interest a noticeable number of people. Our everyday tasks can be solved on much simpler systems that offer an order of magnitude better price/parrot ratio.

Beautiful far away

Well, now it’s time to return to the more or less familiar segment of desktops, laptops and servers. By the way, given the focus of our publication on SOHO (home users and small offices), there is nothing strange in the mention of servers. For example, while relaxing between games in Supreme Commander: Forged Alliance, messages often appear in the local chat that a particular player is using a system based on Intel Xeon or AMD Opteron. From the point of view of cost justification, these players lose; server solutions have always been and will cost more than similar desktop solutions (and the reason, of course, is not that large companies buying servers can pay more). But, at the same time, they appear on the market a month or two earlier than consumer models with similar characteristics, which means that it still makes sense for an extreme gamer or enthusiast to spend extra money to play with the highest settings or simply “be ahead of the curve.” with your benchmarks.

Core i7 – now Intel also has real quad-core processors