Late in 2009, Intel's standing in the high-performance desktop CPU segment was excellent as the company's primary and almost the only rival had just left that part of the market. On the other hand, Intel's positions were threatened in the mainstream segment. Having served many users faithfully for a few years, Core 2 processors became incapable of competing with newer solutions presented by AMD. With just two processing cores, Core 2 Duo products are failing to meet the requirements of the most popular software these days which gradually moves from single/dual-threading to multi-threading. Even despite their high clock rates and very large cache (of the higher-end CPU series). Besides, the latter two features result in higher cost prices which prevent necessary price cuts. As for Core 2 Quad products, they do offer better performance, but they are even more expensive. As a result, in the 50-150-dollar segment, similar performing solutions from AMD can be 30 (or even 50) percent cheaper. On the other hand, similarly priced CPUs from both companies perform very differently. However, Intel wouldn't have been the major market player, if they didn't know how to solve such problems. If older processors are losing the race, then new ones are in order. Thanks to massive investments into R&D (massive even in terms of this industry) and always being the first to implement finer process technologies (which also requires vast expenses, but Intel can afford it), the company has no problems with novelties. In particular, on January 4, Intel officially rolled out essentially new dual-core processors for the socket LGA1156. Let's examine those in detail.

Clarkdale: what's new?

All of the three new series -- Core i5-600, Core i3-500 and Pentium -- are based on the same chip, the further development of the Nehalem. As we have already said, one of the key advantages of the new architecture is its "modular" nature that provides excellent scalability in terms of the number of cores, both upward (from the initial 4 to 6 and 8 in prospective LGA1366 models) and downward. This is a considerable difference from the Core 2 architecture which is strictly dual-core. (Well, there are quad-core and hexa-core processors of that design too, but their cores are connected mechanically, which complicates data exchange between cores and results in other issues.) So, as a CPU, Clarkdale is the example of the downward scalability: it has two cores instead of four, but each is completely identical to those found in Bloomfield and Lynnfield. Against all expectations, Clarkdale even has the same L1 and L2 caches: 64KB and 256KB, respectively. The L3 cache is halved down to 4MB. Since cache occupies a large part of a CPU these days, all these changes are bound to reduce processor cost price. Smaller is cheaper even within the same process technology. However, the latter has changed from 45nm to 32nm. All of these efforts can reduce the cost price by a factor of three. In other words, while Lynnfield can be in processors for $200 and more, Clarkdale can be featured in solutions sold below $100. And it does just that in the form of the new Pentium.

Actually, the reality is a bit worse, because processor cost prices are higher than they could have been. There are two reasons, one temporary, the other permanent. The former is that the initial yield of 32nm CPUs will be lower than the current yield of solutions made according to the polished 45nm process technology. However, as we have said, this is temporary. Besides, the cullage can still be sold -- the Pentium has 3MB L3 cache and lower clock rates not without reason. The latter, permanent reason is that processors actually have dual-crystal design, the second crystal being a graphics accelerator made using the 45nm process technology. On the one hand, adding it to every CPU is dubious, because about one third of the market is occupied by discrete graphics cards, users of which consider the aforementioned integrated graphics an unnecessary expense. On the other hand, the remaining two thirds of the market is still twice as much. Secondly, lots of buyers of discrete graphics also buy more expensive processors. Thirdly, if you look at what major vendors are offering, you will see that even desktops with discrete graphics (not to mention laptops) are usually based on integrated chipsets. Unification of manufacture is quite helpful, you see. This is the reason Intel adds graphics cores to all of its new processors. If those come in handy, good. If not, users still don't overpay that much. To estimate it, just take a look at AMD 785G which is "dirt cheap", but brings profit. (Intel promises the new graphics core will perform similarly. This is a great step forward for the company, because the G45 and earlier products were inferior.) However, the 785G is made using the 55nm process technology, while Intel uses the long-mastered 45nm one which is also better in terms of price. Besides, selling a graphics core with every CPU Intel is going to remain the number one graphics supplier forever.

However, don't expect to find a use for integrated graphics after installing a regular graphics card. Unfortunately, the bad old issue of Intel's integrated chipsets still lives on. The graphics core "locks" PCIe lanes, so you have to choose between it and your usual graphics card. Moreover, if you use the integrated graphics core, you won't be able to use the spare PCIe x16 slot -- it just won't work. On the other hand, at least we can choose now. E.g. if you want to play a game, select the discrete graphics in BIOS and there you go. If you need to browse the Web or do some "e-paperwork", select the integrated graphics and enjoy the silence and power savings. Rebooting every time (as well as re-plugging the cable, if you have a budget monitor with a single video input) is not very comfortable, of course. However, we think this feature will still be used by many desktop owners. It will also come in very handy in laptops. Naturally, to be able to use integrated graphics at all, you'll have to buy a motherboard based on one of the special chipsets with FDI interface support. However, we believe that the H55 and H57 will become mainstream in no time by pushing the Ð55 to the top segment, where one might need two graphics cards or something like that. Anyway, the integrated graphics core is a very interesting subject which we shall examine in a dedicated review.

For now, let's return to the remaining processor units: PCIe, DMI and memory controllers. There's nothing much to describe here, because they are identical to all other controllers found in LGA1156 processors. The former supports the same 16 PCIe 2.0 lanes. If a P55-based motherboard is used, those can be separated for two slots for multi-GPU configurations and whatnot. Naturally, the DMI interface is still there as well, providing forward and backward compatibility with motherboards. Finally, dual and quad-core Core i5 processors have similar memory subsystem features: DDR3-1333 support (unlike Core i7, the official clock rate is also the maximum, because there's no multiplier of 12) and the UnCore clock rate of 2.13 GHz. The Core i3 also support DDR3-1333 memory, but they may even have a lower UnCore multiplier. The Pentium is limited to DDR3-1066, which doesn't actually matter much in terms of performance.

Now let's say a few words on new processors' market positioning. As we assumed about six months ago, from this year forward, processors with 6MB and 2x6MB L2 (Å8x00, Q9x50) will be disappearing from Intel's roadmaps due to the insufficient price/performance ratio. Those will still be on sale for quite some time though. However, processors featuring cores with 2MB and 3MB L2 (not to mention the Celeron with the 1MB cache), as well as products based on those, will have a longer lifespan. These solutions will be the ones to compete with today's novelties in the minds of Intel's target audience. Which against which exactly? The Core i5-670 is not just the top-end dual-core product, the company places it just slightly below the Core i7 thanks to the very high clock rate (which benefits single/dual-threaded applications) and low TDP. Correspondingly, the rollout of Core i7-860S, Core i5-750S and the 600 series will let the company discontinue all energy-efficient modifications of Core 2 Quad. From this year forward, all willing to build a silent, cold PC should only consider solutions for the socket LGA1156.

The Core i5-660 and 661 is placed between Core i5-750 and Core 2 Quad Q9505: slightly below the former and slightly above the latter (despite the somewhat lower initial price). The Core i5-650 should join the serried Core 2 Quad Q8000 series: just below the Q8400, but above the Q8300 and Q8200. In other words, despite having just two cores, the new processors are called to free the market from the older quad-core models. Let's see if Hyper-Threading will help them do it. However, we are sure that the Core i3-540 and 530 will cope with their task easily: the former is placed just above the Core 2 Duo E7600, the latter between the E7600 and Å7500. The similar clock rates will provide parity in fewer-threaded applications, while Hyper-Threading should do the trick in multi-threaded ones, because we're talking "2 cores vs. 2 cores + HT" here, not "4 cores vs. 2 cores + HT". Oh, and Pentium naturally fits among siblings: somewhere around the E6300, slightly below the E6600 for the time being.

Testbeds

The main character of this review is the Core i5-661. Though consider we have tested the Core i5-660 as well. These two processors differ by graphics core clock rates, TDP and supported virtualization technologies, but their CPU core perform identically. Of course, we'd like to present more test results (in particular, the cheaper Core i3 series), but obtaining novelties before their official rollout can be hard at times, especially if holidays are involved. Anyway, this isn't our last review of the LGA1156 platform, so, hopefully, we'll examine all the corresponding processors eventually.

Speaking of competitors, the Core 2 Duo E8600, unsurpassed in its class, comes to mind at once. Moreover, it has the same initial clock rate of 3.33 GHz. However, as we have found out before, a fast dual-core CPU might not be generally fast in today's software, because there are lots of applications in which budget quad-core models perform far better. Let's see if the Core i5-661 can cope with the significantly cheaper Athlon II X4 630. Of course, we couldn't pass the Core 2 Quad Q9505 and Core i5-750 over, as Intel places the Core i5-661 right in between. And since we're examining the upper part of the mainstream segment today (Core i5-660/661 and Core i5-750 cost about $200, while the Core 2 Duo E8600 and Quad Q9505 are even more expensive), we should also add the Phenom II X4 965, AMD's fastest. The latter has higher TDP, but it should provide decent performance as well. Besides, it has the same amount of cache as the Core 2 Duo E8600 (though it's slower), as well as the clock rate of 3.4 GHz similar to 3.3 GHz of the Å8600 and 3.33-3.6 GHz of the Core i5-660/661. By the way, the Core i5-750 is also close by thanks to Turbo Boost -- if only one or two cores are used, the clock rate can be increased from the initial 2.66 GHz to 3.2 GHz.

In other words, we have a very representative group here that completely covers the market segment of the Core i5-661. We also have a representative of the cheaper segment. We need it, because if Core i5-66x only catches up with or lags behind Athlon II X4 630 in some tests, it will mean that the score of the cheaper Core i5 processors and that of the entire Core i3 series (called to compete with the Athlon II) is even more deplorable. Those willing to compare the novelties with other processors from Intel and AMD can browse our traditional summary table.

Write a comment below. No registration needed!