Today we are going to review the smallest of the new enthusiastic range of intel, which is none other than the i7 5820K. This is the first time that 6 cores have been included in the cheapest processor model, compared to the 4 that both the i7 4820K and i7 3820 exhibited, both belonging to the original 2011 socket.

The counterpart in this case to justify the jump to higher models is none other than the number of pciexpress lanes. In socket 2011 all the processors mounted 40 pciexpress lanes, in socket 2011-3 only the two highest models incorporate that large number, leaving us with more than respectable 28 pciexpress lanes in the processor in question.

It is a processor with Haswell-E architecture, manufactured at 22nm, and with 6 cores, which makes it a very similar processor to the high-end of the previous generation, with improvements in terms of consistent but little IPC, just like that we already saw in the jump from socket 1155 to 1150. The great innovations come from the chipset, completely renewing the platform, eliminating backward compatibility and incorporating for the first time a DDR4 memory controller in consumer equipment. Let's see in detail what news this platform brings us.

Product sponsored by:

What's new on the X99 platform

Intel has taken much advantage of its old enthusiastic X79 platform, launched almost 3 years ago alongside the i7 3960X. And although in its day it was top of the range in everything, it began to show little, after all the years go by, and in hardware 2 generations of processors and these 3 years have been more than enough to leave it far behind its smaller but more modern ranges.

DDR4 memory is introduced for the first time, along with Intel's first home-based 8-core processor. As with every socket change, expectations are high on this platform.

On a physical level, we find very few changes, good news for those who have socket 2011 heatsinks, since the anchors are identical and are fully compatible. A slightly larger than normal notch in the center of the socket prevents us from pricking an old 2011 socket processor into the new boards.

Important changes come if we delve a little deeper. We recall that little by little, more functions of the chipset are integrated into the CPU, from the first Sandy bridge the chipset of a board is nothing more than a somewhat oversized Southbridge, dealing with the "low" speed expansion ports and little more. In this case we continue in the same line, with light but constant advances, which are not much more than just to bring this enthusiastic platform to the level we have already seen in socket 1150.

In this field we see that the X99 chipset brings many improvements compared to its predecessor, with 6 USB3.0 ports and 8 USB2.0 ports, a Gigabit Ethernet port, HD Audio, and the typical 8 lanes pciexpress 2.0 for expansion cards. The most obvious improvement is probably the inclusion of a whopping 10 SATA3 ports. We see the choice of the DMI 2.0 connection to communicate the processor and the chipset, used in socket 2011 as a bit risky, since with “only” 20Gbit / sec it could be a considerable bottleneck in the case of simultaneous use of many of integrated connections, especially from many SATA ports with fast SSDs. It's not a regular scenario, but it's not outrageously rare on an enthusiastic platform either.

Another big change is the support for DDR4. Many wonder if it really makes sense, since memory bandwidth today is not a major limitation on a personal computer with the affordable DDR3 kits of 2133-2400MT / s, even on dual channel platforms.

The answer is no, not much at least. The biggest reason for the inclusion of DDR4 is to pave the way for the business market, servers with ECC memory and many times hundreds of Gigabytes of RAM, where not only extra bandwidth will be welcome (remember that in socket 2011 the fastest memory officially supported is 1866MT / s), but also the energy savings that will bring with kits running at 1.2V. The first memory kits that we will see for the consumer market usually use somewhat higher voltages (1.35V seems to be the choice of many brands), but it is also a significant efficiency jump compared to the 1.5-1.65V of similar DDR3 kits. frequencies, and we also remember that the consumption of the memory subsystem in a personal computer, with the consumption of graphics, CPU and hard drives, is minimal. However, in server racks, every watt is valuable in the long run.

Technical characteristics of socket processors 2011-3

specs

i7-5820k in detail

We see the classic packaging from Intel for its high-end, we note that they have used their already typical blue color scheme on the box, unlike the black scheme of the Extreme Editions

i7-5820K

i7-5820K rear box

i7-5820K box front

Inside the box we find the manual and the processor, correctly protected but without any additional extras.

Once again, since it is an enthusiastic range processor, it does not include a heatsink as standard, a habit that Intel started with its Sandy Bridge-E, again leaving the user the choice regarding cooling. It is not a great drawback, in fact it is an advantage, since it eliminates unnecessary costs and are processors that generate enough heat and that would also make a quality cooling system recommended. If the 2011 socket processors were already large, this is even more, with its 356mm ² die size, it leaves the 257mm ² of the Ivy-E hexacores small (although it does not reach the values of Sandy Bridge-E).

Intel has not opted for differentiated models this time, since they all start from the same wafers with 8 cores, only that the two simplest processors have two of them disabled. The disabled core pair is not always the same, although they must be two cores located in the same row (oriented according to the diagram in the previous section).

As we anticipated at the beginning of this review, we are faced with a 6-core processor with hyperthreading technology (that is, it appears before the OS as 12 process threads), with support for 4 channels of DDR4 memory (compared to 2 of the platforms socket 1150/1155), a respectable 28 PCI Express 3.0 lanes (compared to 16 + 4 for the minor sockets and 40 for the 5930K and 5960X) and Haswell architecture.

This pciexpress line configuration is quite convenient, more than enough for a single GPU, very suitable for 2 graphics cards (even if we have more pciexpress cards, such as sound cards), and sufficient even for 3 (operating at 8x / 8x / 8x). We do not recommend mounting 4 graphics cards with this processor, as the lack of lanes begins to be evident.

On the chipset we have 8 lanes pciexpress 2.0, which share bandwidth with the M.2 slot for SSD that many boards incorporate. Today in this range it is somewhat scarce for graphics, but it is an ideal addition for pciexpress expansion cards, especially in cases like this 5820K that the native controller does not have lines to spare precisely.

Although the 4930K was the option to consider on the X79 platform, I would say that in this it is not the 5930K that takes its place, but this great i7 5820K. For a price that does not reach € 400 we find a processor with 6 cores, 15Mb of L3 cache compared to 20 of the 5960X and… that's all. There the disadvantages end. The base frequency is somewhat lower than in the Ivy-E, but this is not at all a problem with a fully unlocked processor, which with a little overclocking will easily outperform its older brother, the rather more expensive 5930K, whose only advantage are the 40 lanes that we mentioned.

Due to its more conservative frequencies, in applications that do not take advantage of all the cores we see in the i7 4790K a tough contender to beat, which surely will remain for many the reference processor when it comes to video game performance, but we do not doubt that this trend It will change, facing the future, the 5820K with its 6 cores is a much safer bet, as was the 775 quad core cores, which also performed worse than the equivalent dual cores, which exhibited much higher frequencies that gave them the advantage, and over the years they became much more competent.

The TDP increases to 140W, from a generous 125W to the 4930K. In our tests, the measured consumption has been similar or slightly lower than its predecessors, so we imagine that Intel simply wants to play it safe and not leave its hottest (although also much more powerful) 5960X in the wrong place.

In this case, the jump to DDR4 memory carries official support for memory frequencies that had been between us for a long time. They have also become more demanding in terms of BMI voltage, with even lower tolerances than usual.

Testing equipment and performance tests

TESTING BENCH
Processor:	Intel i7 5820K
Base plate:	Asus Rampage V Extreme
Memory:	Crucial DDR4 4x8gb 2133MT / S CL15
Heatsink	Cooler master Seidon 120XL + NB Eloop 1900rpm
HDD	Intel X-25M G2 160Gb
Graphic card	Asus 780Ti Matrix Platinum
Power supply	Antec High Current Pro 850W

Synthetic tests

We start the benchmark stack with a multi-threaded test that is an excellent representative of the overall performance of the CPU / RAM suite, the well-known Cinebench, based on Maxon's Cinema 4D software.

Although at stock frequencies this processor does not reach the values of the 4930K, partly due to the slower RAM, partly due to the more conservative turboboost frequencies, it has no problem to stand out and lead the table, at least while waiting that the 5960X will dispute its position. A very favorable test for processors with a lot of gross power, as such stands out clearly from the quad core and leaves the small pentium G3258 in the gutter. In summary, if we are going to dedicate ourselves to multithreading, image rendering, video editing, the 5820K is the way to go, unless we want to invest in its older brother.

The POV-Ray test, a raytracing software that takes advantage of all available threads, makes even more evident the improvements that the Haswell architecture has brought. It is a very favorable scenario, and it is clearly seen that clock a clock has earned a good pinch, showing itself to be superior to the already excellent 4930K both in stock both and overclocking both.

In the 7-zip benchmark we also see high performance values, although in this case there seems to be a slight regression from the 4930K's performance, in this case the haswell improvements are not used intensively and the slight drop in frequencies makes the rest. This test uses the LZMA compression algorithm, and is also a benchmark that makes the most of all available threads, also being a true reflection of the performance that we can expect by compressing and decompressing files with any modern software. It should be noted that WinRar, although in its previous versions it was also limited to 1-2 cores, right now follows the same trend.

Games Tests

3DMark is probably the best when it comes to evaluating a team's gaming performance at a glance. It is a synthetic test, and as such it is not exempt from a certain controversy about its objectivity, but it is clear that it is a very good indicator of what we can expect from a team. We have used the Fire Strike test, which is the most comparable to the demands of the latest generation titles.

As we expected, it is the performance of the graph that is most decisive here. Even with an i5, the overall result would not have suffered too much. However, you can see a really good scaling in the result of physics, where a processor like the i7 5820K does stand out with the authority of cheaper options. This processor is slightly below the 4930K that we analyzed, in this test, although with overclock it recovers the difference without problems. As we mentioned previously, for the calculation of physics the 6 cores are really recommended, in fact this result will be the terrain of the 5960X, although as we anticipated, the global score is not expected to rise significantly.

WE RECOMMEND YOU Asus RX VEGA 64 Strix Gaming Review in Spanish (Full Review)

In real games, we see that the trend seen in 3DMark is maintained: The bottleneck in high-end equipment is still the graphic power. Even with one of the best custom models of the most powerful monogpu on the market, the limitation is still the graphics, as we see hardly any overclocking of the CPU, both the 4930K and the 5820K are more than enough to play any title at today.

We note that this is not the case for all titles, since in Crysis 3, or in the larger multiplayer maps of Battlefield 4, there is a very clear gain with processors with more cores, and a processor as powerful and popular as it is the i5 2500K can reach 100% use depending on which cases. At the moment, these cases are a minority, but it is to be expected that in the coming years the exception will become the norm, and it will be more and more common to see games in which these hexacores are increasingly unmarked, despite their lower frequencies..

Consumption and temperatures

As always with the most powerful processors, we expect to see consumption values in the highest range of the tables. Haswell brought great efficiency improvements on laptops, let's see if they have been kept on these high-end processors.

The values under load are as expected, it is surprising to see that this processor, even with 140W of TDP, has a consumption almost identical to the 4930K for which 125W is specified. Being an unrealistic test like Linpack, it is a pleasant surprise to see consumption below the TDP, since it is common for it to be exceeded by far, which confirms what we already intuited, the 5820K is a very powerful processor, of moderate consumption, but also very efficient. Equally, I would have liked to see a bigger jump compared to the last generation. We will have to wait for the jump at 14nm to start seeing movement in this area. Although the improvements are slight, the consumption in idle improves an already excellent value of its predecessors, together with the DDR4 RAM, it can mean a very slight saving, but appreciable in long times.

Overclocking, of course, brings a considerable increase in consumption. The loss in efficiency is not as severe as it might seem, since although the processor consumes more electricity, it also performs more operations at the same time, completing the same task in slightly less time than without overclocking. Unfortunately we cannot test this processor with the liquid used that we used in the review of the 4930K, so we will use a simple radiator kit, with the caveat that the results cannot be compared as is.

Looking at the temperatures and observing this difference, although it is not about fresh processors, we see that the improvement in terms of temperatures is evident, since with a kit that has considerably less cooling power, we see similar temperatures, in fact much lower than stock, which we saw in the 4930K. Our congratulations to Intel, which looks like it's finally starting to polish up those heat transfer issues to the IHS that make processors as efficient as the Ivy Bridge-E, even with soldered cores, hotter than desirable.

conclusion

The Intel Core i7 5820K is currently the third most powerful processor that can be mounted on a home computer, and without a doubt an excellent choice in terms of quality / price within the high range. It is a pity that the DDR4 memories and the boards penalize the price of the platform so much, because otherwise we would have for 30 € more than what an i7 4790K is worth, a processor that beats it in any multithreaded task.

It is not an economic processor, although luckily it is in a much lower price range than the hexacores of past generations. Again, the gaming gain versus the 4930K is not particularly noticeable, and the limitation is still the graphics in most titles, so the biggest reason to upgrade would be because of the new features of the X99 chipset over the X79 rather than by the processor.

In summary, this processor is the best option if we want to assemble a new team oriented to heavy tasks at a moderate price, without neglecting games or 1-thread performance. In case of having more budget for the team, the 5960X gains strength as the most powerful alternative today, that is, almost 3 times more expensive than the processor in question. We hope to be able to publish the review of this other processor in the coming weeks.

ADVANTAGE	DISADVANTAGES
+ VERY GOOD PERFORMANCE, AS MULTI-WIRE AS 1 THREAD	- LITTLE IMPROVEMENTS ABOUT ITS PREDECESSORS THAT JUSTIFY THE UPDATE
+ WELDED CORES TO THE IHS, TO IMPROVE TEMPERATURES AND FACILITATE OVERCLOCK	- SUFFICIENT PCIEXPRESS LANES FOR GENERAL USE BUT SCARCE FOR THE MOST EXTREME MULTIGPU SETUPS
+ OVERCLOCK CAPACITY, SUPPORT FOR MULTIPLE BCLK AND UNLOCKED MULTIPLIER	- ALTHOUGH IT IS NOT A PROCESSOR FAILURE LIKE SUCH, DDR4 RAM VERY EXPENSIVE AND HARD TO GET IN THESE MOMENTS
+ MEASURED CONSUMPTION FOR THE POWER OF THE PROCESSOR. CONSUMPTION IN LOW IDLE.
+ THE NEW X99 PLATFORM, FINALLY THE ENTHUSIASTIC RANGE IS UPDATED
+ MODERATE PRICE, BUT LESS THAN USUAL FOR HEXACORES