Performance comparison table for intel and amd processors. Comparison of performance of different generations of Intel processors

62 processors and 80 different configurations

Another year has changed on the calendar, we have prepared new testing methods computer systems, which means that it’s time to sum up the results of processor testing (which is a special case of system testing) in 2015. Last year's results were quite brief - they included the results of only 36 systems, differing only in processors and obtained exclusively using the GPU built into them. This approach, for obvious reasons, left behind a considerable number of platforms that lack integrated graphics, so we decided to modify it a little by sometimes starting to use a discrete video card - at least where it is needed. However, the 2015 tests became to some extent “educational and training” - in 2016 we plan to further refine the approach to testing in order to further bring it closer to real life. But be that as it may, today we will present the results of 62 processors (more precisely, there are 61 different ones, but thanks to cTDP, one of them is worth two). And that's not all: 14 of them were tested with two “video cards” - an integrated GPU (different for everyone) and a discrete Radeon R7 260X. We also tested four processors for the latest LGA1151 platform with two types of memory: DDR4-2133 and DDR3-1600. Thus, the total number of configurations was 80 - this is much less than 149 in the results before last, but for those we collected information for two and a half years, and the “lifetime” of the current test method was approximately eight months, i.e. almost three times less. In addition, the unification of tests for different systems allows you to compare the results with those obtained when testing laptops, all-in-one PCs and other complete systems.

But in this particular article, as mentioned above, we will limit ourselves to processors. More precisely, systems that differ mainly only in processors - it is clear that there is no other meaning to “testing processors” (especially for different platforms) has not had for a long time, although for some it is still a revelation :)

Test bench configuration

Since there are many subjects, it is not possible to describe their characteristics in detail. After thinking a little, we decided to abandon the usual short table: it still becomes too vast, and at the request of the workers, we still included some parameters directly on the diagrams. In particular, since some people are asking to indicate right there the number of cores/modules and computational threads running simultaneously, as well as operating clock frequency ranges, we tried to do just that. If readers like the result, we will save it for other tests in the coming year. The format is simple: “cores/threads; minimum/maximum core clock speed in GHz.”

Well, all other characteristics will have to be looked at in other places - the easiest way is from manufacturers, and prices - in stores. Moreover, for some devices the prices are still indeterminable, since these processors themselves are not available in retail (all BGA models, for example). However, all this information is, of course, also in review articles devoted to these models, and today we are engaged in a slightly different task than the actual study of processors: we collect all the data obtained together and look at the resulting patterns. Including paying attention to the relative position not of processors, but of entire platforms that include them. Because of this, the data in the diagrams is grouped precisely by platform.

Therefore, all that remains is to say a few words about the environment. As for memory, the fastest one supported by the specification was almost always used. There are two exceptions: what we called “Intel LGA1151 (DDR3)” and Core i5-3427U. For the second, there were simply no suitable DDR3-1600 modules, so it had to be tested with DDR3-1333, and the first - processors for LGA1151, but paired with DDR3-1600, and not the faster (and “main” according to the specifications) DDR4-2133 . The amount of memory in most cases is the same - 8 GB, with the exception of two versions of LGA2011 - here there were 16 GB DDR3 or DDR4, respectively, since the four-channel controller directly provokes the use of a larger amount of RAM. The system drive (Toshiba THNSNH256GMCT with a capacity of 256 GB) is the same for all subjects. As for the video part, everything has already been said above: discrete Radeon R7 260X and built-in video core. The video core was always used when the processor had one (with the exception of the Core i5-655K, since the first version of Intel HD Graphics is no longer supported by modern operating systems), while a discrete video card was used where there was no built-in video. And in some cases - where there is an embedded video: to compare the results.

Testing methodology

To evaluate performance, we used our performance measurement methodology using a benchmark. We normalized all test results relative to the results of the reference system, which last year was the same for laptops and all other computers, in order to make it easier for readers to make the hard work of comparison and selection.

Thus, these normalized results can be compared with those obtained in the same version of the benchmark for other systems (for example, we take it and compare it with desktop platforms). For those who are interested in absolute results, we offer them as a file in Microsoft Excel format.

Video conversion and video processing

As we have noted more than once, in this group a discrete video card allows you to increase performance, but this effect is clearly visible only on older platforms (such as LGA1155), where the power of the integrated GPUs itself was small. Actually, this is the answer - why did they increase it in new generations: and so that there would be no incentive to buy a video card as well :)

The dependence of performance on the number of threads of executed code is also clearly visible here. As a result, we come to a very wide range of results - they differ by more than an order of magnitude, since low-end dual- and quad-core CULV solutions (such as the old Celeron 1037U or the slightly newer, but also outdated Pentium J2900) give only ≈55 points, and the top eight-core Core i7-5960X - all 577. But the main “crush” is unfolding in the mass segment (up to $200): modern Core i5s can increase productivity (relative to the “floor level”) five times, but further investments only double it. Actually, there is nothing surprising in this: the higher, the more expensive.

As for comparing platforms, then... they don’t need to be compared. Indeed: desktop AMD FM2+ roughly corresponds only to Intel ultrabook processors, and formally, the top-end AM3+ only corresponds to the long-outdated LGA1155. However, Intel's growth from generation to generation is small - even in such well-optimized tasks we can only talk about 15-20% at each step. (This, however, sometimes leads to qualitative changes - for example, the Core i7-6700K has actually caught up with the once top six-core i7-4960X, despite a significantly lower price and a simpler device.) In general, it is clear that manufacturers are dealing with completely different issues , and not at all attempts to greatly increase the performance of desktop systems.

Video Content Creation

As we have already written more than once, in this group the multi-threaded test in Adobe After Effects CC 2014.1.1 gave us a real blow. For him normal operation It is recommended to have at least 2 GB for each calculation thread - otherwise the test may “fall out” into single-threaded mode and start running even slower than without using Multiprocessing technology (as Adobe calls it). In general, for full-fledged work eight threads require 16 GB of RAM, and an eight-core processor with NT will require a minimum of 32 GB of memory. On most systems, we use 8 GB of memory, which is enough for “eight-stream” systems when using integrated video (if they have it: for Desktop Core i7 does this, but the FX-8000, for example, does it worse), but not discretely. Another stone in the garden of those who still believe in “processor testing” as something independent - in isolation from the platform and other environment: as we see, sometimes attempts to make it equal lead to extremely interesting effects. A “pure” comparison is perhaps only possible within the same platform, and even then not always: the amount of memory required by some programs may depend on the processor itself and not only it. Which just hits the top models hard, because they need more, and “more” in this case means more expensive.

However, in any case, in this group of applications the “processor dependence” is less pronounced than in the previous one - there the older Core i5 outperformed the low-voltage surrogates by five times, and here only a little more than four. In addition, a more powerful video card can increase the results noticeably less, although it should not be neglected (if possible) either.

Digital Photo Processing

This group is interesting in that it is absolutely different from the previous ones - in particular, the degree of “multi-threading utilization” is much lower here, which noticeably reduces the range of results obtained, but here are the differences between the Core i5 (we will continue to be tied to this family as upper level mass segment - sales of systems based on more expensive processors are incomparably lower) and entry-level devices exceed six times. What is this connected with? Firstly, there is a noticeable dependence of performance on the GPU. First of all, integrated: discrete cannot develop to its full potential due to the need for frequent data transfers. But the power of integrated graphics in low-end and high-end processors differs significantly! And we should not forget that not only quantitative, but also qualitative differences between junior and senior processors still remain - for example, in terms of supported instruction sets. This hits hard both on the younger Intel families (remember that Pentium, for example, still does not support AVX) and on outdated processors from both companies.

Vector graphics

But here is an illustrative example of what modern software can be different. Even if we are talking about, to put it mildly, not the cheapest programs, and not for “home use”. In fact, as we have noted more than once, any serious optimization of Illustrator was last made about 10 years ago, so the program for fast work we need processors that are as similar as possible to the Core 2 Duo: a maximum of a couple of cores with maximum single-threaded performance and without support for new instruction sets. As a result, they look the most advantageous (taking into account the price) modern Pentiums, and processors are more high class may be faster than them only due to a higher clock speed. Processors of other architectures feel very bad under such conditions. Actually, even in the Intel line, such intensive methods of increasing performance as adding a fourth level cache, in this case only hinder, not help. However, in any case, trying to greatly speed up work in this program (and similar ones) is not a very promising endeavor: only a fourfold difference between the best Core i5 and surrogate platforms speaks for itself.

Audio processing

Here is an example of a situation where, it seems, the computational cores are not superfluous, and even the GPU matters, etc., but the difference between the Celeron N3150 (the slowest in this test) and the Core i7 for mass platforms is only about five times . Moreover, a considerable part of it can be attributed to the surrogacy of younger architectures - the very old Celeron 1037U (albeit very limited, but a full-fledged Core) is almost one and a half times faster than the N3150, and the younger desktop Pentiums are three times faster. But further... the more expensive it is, the less effective the amount of the “additional payment for the processor” is. Even within the same architecture - AMD's "construction equipment" with its "budget multi-threading" in this case is only able to compete with the same Pentium: six threads are faster than four from the same manufacturer, but do not look convincing against the background of only two cores from a competing design.

Text recognition

Not at all like in the previous case - here the FX-8000 still easily outperforms any Core i5. Note that AMD positioned them this way at the time of release: between i5 and i7. Including the price. Which, unfortunately, later had to be radically reduced, since the number of such “convenient” tasks turned out to be not too large. However, if the user is specifically interested in them, this provides an opportunity to save a lot of money. Considering, of course, that this family has not been updated for more than three years (in a serious way, in any case), and Intel processors are slowly but growing.

And the problem of scalability is also clearly visible - no matter how good the additional cores and threads are, the more there are, the less effect the increase in number gives. Actually, in the end, you shouldn’t be surprised that this process stopped long ago in mass-produced processors - we need even more convincing arguments for multi-cores than can still be found. Here are four modern cores - good. Four dual-threaded cores are even better. And then that’s it.

Archiving and unarchiving data

If archiving uses all cores (and additional computing threads) of processors, then the reverse process is single-threaded. Considering that it has to be used more often, this could be considered a nuisance if the process itself were not significantly faster. Yes, in fact, packaging has become a fairly simple operation to pay close attention to when choosing a processor. In any case, this is true for mass-produced desktop models - low-power specialized platforms can still “tinker” with such tasks for a long time.

Speed ​​of installation and uninstallation of applications

In principle, we introduced this task into the test methodology mainly due to the need to test ready-made systems: And on the same processor in different environments, as we already know, performance can differ by one and a half to two times. But when the system uses a fast drive and enough memory, the processors themselves do not differ fundamentally from each other. However, surrogate platforms may well turn out to be exactly two to three times slower than “normal” desktop ones. But the latter differ little from each other - be it Pentium or Core i7. Essentially, all that may be needed from a processor is one thread of calculations with maximum performance. But mobile systems aside, this is almost always done to about the same extent.

File operations

And these are especially “platform-cumulative” tests rather than processor tests. As part of this line of tests, we use the same drive - with all that it implies. But the “platform” can matter - for example, the results of LGA1156 turned out to be a bit of a surprise: it seems not the worst desktop solution, which until recently could be considered even fast (the LGA775 still found among users is even worse), but it turned out that under such loads it can only be compared with Bay Trail or Braswell. And even then, the comparison will not be in favor of the “old lady” who was once close to the top level. But modern budget systems are practically no different from non-budget systems - simply because the former are already enough for performance to begin to be determined by other components of the system, without being limited by the processor or even the chipset.

Total

In principle, we made the main conclusions about processor families directly in the reviews, so they are not required in this article - this is primarily a generalization of all previously obtained information, nothing more. And generalizations, as we see, can sometimes turn out to be interesting. Firstly, it is easy to notice that the influence of discrete video cards on performance in mass-produced programs can, in general, be considered absent. More precisely, in some applications it is, but being “spread” across all tests, it quietly and peacefully evaporates. In any case, this is true for more or less modern platforms - it’s easy to see that weak integrated graphics from the LGA1155 era, even overall, can reduce the results by five percent, which is more or less noticeable, although not critical. The same should apply to older discrete video cards, which will also be inferior to slightly newer ones, but in this case, the border between “good” and “bad” solutions is no longer pushed back by three, but by five or more years from the current moment. In short, modern platforms are free of such problems. So, for a quality comparison, it is not at all necessary to require the same video part, which means that if you need, for example, to compare a laptop with a desktop system, we find a suitable article about a laptop (not even necessarily about the same one - another one on a similar platform will do) and compare. The data storage system is even more important, so if there is no parity in the articles on it, you will have to limit yourself to the results of groups of tests that do not depend on the drive. As for the video... We repeat: among mass applications not so strongly attached to him, but gaming application- a completely different story.

Now let's try (as usual) to look at the range of performance that we managed to cover this year. The minimum result in the overall standings is the Celeron N3150: 54.6 points. The maximum is for the Core i7-6700K: 258.4 points. “Professional” platforms like LGA2011/2011-3 failed to take first place, although in some tests its “multi-core” representatives were confidently in the lead. The reasons for this have been voiced more than once: manufacturers of mass software mainly focus on the fleet of equipment available to users, and not at all on some “shiny peaks”. There are (and always have been and always will be) tasks for which computing resources are “always in short supply”, and it is for them that top-end systems are required (sometimes going far beyond the scope of our testing), but the bulk of problems can be easily solved on a mass-produced computer. Often even outdated.

In this regard, it is interesting to compare the current “Results” not with the past ones, but with the ones before last. Then testing was done according to a completely different scheme - always using a powerful discrete video card. And there were more professional applications, so the top six-core processors, in general, still turned out to be faster than the best solutions for mass platforms. However, at the same time, the Core i7-4770K scored 242 points - which is just comparable to 258.4 for the Core i7-6700K (from the point of view of time-adjusted positioning, these processors are the same: one was the fastest solution for the mass LGA1150 of 2013, and the second - the same in 2016 for LGA1151). At the same time, both then and now, various Pentium/Core i3/Core i5 were pushed in the range of 100-200 points - nothing has changed. Except that the scores have changed: the software was mentioned above, but the standard has also changed. Previously, this was the AMD Athlon II X4 620 (budget, but desktop and quad-core processor) with a discrete Nvidia-based graphics card GeForce GTX 570. And now this is an (ultrabook) Intel Core i5-3317U without any discrete data. It seems like everything is different. But in practice it’s the same: a budget desktop gives a hundred points, any investment in it best case scenario can increase productivity (on average across task classes) by two and a half times, and a compact nettop on a surrogate platform will work two to three times slower. This is the state of affairs in the segment desktop computers has been established and has been preserved for a long time, which is clearly shown by our summary results. In general, when going to the store to buy a new computer, you don’t need to read any articles - just analyze the amount of money in your wallet :)

When are tests still needed? Basically - when the task arises to change old computer to a new one. Especially when it is planned to “move to another class”: by changing the desktop to a nettop or laptop, for example. When purchasing a new solution of the same class, you don’t have to worry: the new Core i5, for example, will always be faster than the old one of the same class, so there is no great need for accurate estimates of “by how much.” But the fact that the performance of processors for various purposes is slowly but surely growing can lead to pleasant surprises - when, for example, it turns out that an old desktop can easily replace an ultrabook, and without any negative consequences. Well, as we see, this is quite possible, since everyone “grows”.

Intel is one of the two most popular companies developing processors for laptops and computers. Many gamers and other users consider this company to be the best and prefer its products. But Intel has a fairly wide range of models. So, figuring out which processor is best for which computer is sometimes not so easy. To make it easier for customers to navigate a wide range of offers from the manufacturer, we have created a rating Intel processors. With it you can easily choose the processor to suit your taste.

No. 10 – Intel Pentium G4400

Price: 5745 rubles

And our top chipset begins, called Intel Pentium The G4400 is an excellent option for budget personal computers.

At the core of this processor It is based on the Skylake architecture, it consists of two cores, the clock frequency of which is 3.3 GHz. Additional performance of the device is provided by cache memory, the volume of which here is 3072 KB.

The Pentium G4400 is also capable of image processing. There is a built-in SkylakeIntel HD Graphics 510 graphics processor. Of course, it cannot completely replace a full-fledged video card, but it is enough to perform simple tasks.

This model has a special controller that supports two-way data transfer between the processor and RAM.

This controller is capable of working with memory modules up to 64 GB. So there shouldn’t be any problems installing the required amount of RAM.

Intel Pentium G4400

No. 9 – Intel Pentium G4620

Price: 7085 rubles

Intel Pentium G4620 is a dual-core processor with clock frequency 3700 MHz. It is developed using 14nm process technology. The basis of this device is the Kaby Lake architecture.

This model has the same cache memory - 3 MB, but the graphics processor here is slightly more powerful than the HD Graphics 630. Of course, if we compare the Pentium G4400 and G4620, then the latter option is better, but not much. It is unlikely that you will notice a significant difference in performance.

Nevertheless, the G4620 is an excellent processor, which, of course, is not suitable for professional gamers, but may well satisfy the needs of the average user or lover of playing old games.

In general, it will cope with new games, but there will be slowdowns, and it won’t be possible to set the settings to maximum. If this is not a problem for you, then the G4620 is worth taking. Otherwise, it’s better to take a closer look at more expensive models.

Intel Pentium G4620

No. 8 – Intel Core i3-8300

Price: 12955 rubles

Having finished with the budget segment, let's move on to entry-level processors. Intel Core i3-8300 is already a quad-core processor with a clock frequency of 3.7 GHz. The cache memory here is also twice as large – as much as 8 MB.

The Core i3-8300 comes with an excellent cooler, which is actually rare for powerful processors. Usually, when you buy a really good processor, you definitely need to buy a cooling system for it, because the basic one, as a rule, is terribly insufficient to maintain normal operating condition. But in this case, the boxed cooler copes with its task quite well.

The Core i3-8300 is a good processor that, in conjunction with an equally good video card, can handle most modern games.

In addition, it is sold for a meager price, considering all its advantages. So, if you do not need the most powerful, but high-quality chipset, we recommend choosing the i3-8300.

Intel Core i3-8300

No. 7 – Intel Core i3-8350K

Price: 13100 rubles

Intel Core i3-8350K is an improved version of the previous model. Like the basic version, it has four cores and 8 MB of cache, but its clock speed is 4 GHz.

This is a fairly high figure that is guaranteed to provide you with high performance. The main advantage of the Core i3-8350K over the Core i3-8300 is the unlocked multiplier.

That is, the processor can also be overclocked. Thus, the already high clock frequency of 4 GHz can be increased to 4.6 GHz. This is a pretty good overclocking for Intel processors.

The Intel Core i3-8350K maintains adequate temperatures well. At active work with a computer you are unlikely to heat it above 50 degrees, which is simply an excellent indicator.

Without a doubt, in the Intel model table, this is one of the best processors in terms of price and quality.

Intel Core i3-8350K

No. 6 – Intel Core i5-8400

Price: 16575 rubles

The golden mean in model range companies occupy Core i5 chipsets. It includes fairly current, but still affordable processors. We'll start our review with the Intel Core i5-8400.

This is a six-core processor with a clock speed of only 2.8 GHz, but this is only in standard mode. In turbo boost, when maximum performance is required, it accelerates to 4 GHz. The cache memory here is 9 MB.

The i5-8400 processor is quite popular, because it has six ultra-fast cores and is sold at a very decent price compared to older models.

Overall, this is a more than decent processor. The only downside is that it has sudden temperature fluctuations, but usually it does not heat up above 61 degrees. This model is more than enough for any modern games.

Intel Core i5-8400

No. 5 – Intel Core i5-8600

Price: 18990 rubles

Improved hexa-core fifth processor Intel generation The Core i5-8600 has a significantly higher clock speed. The base frequency is 3.1 GHz, but in turbo mode this figure increases to 4.3 GHz. Otherwise, the technical specifications are the same.

The undoubted advantage of the Core i5-8600 is that in some cases its performance can be equal to even the newest processor models from Intel.

There is also very little heat generation, which is quite good for such a powerful chip. In short, the i5-8600 is an excellent representative of the mid-price segment that will provide you maximum performance even in new games.

Intel Core i5-8600

No. 4 – Intel Core i5-9600K

Price: 21,750 rubles

The Intel Core i5-9600K, being the most advanced model in the line, has again advanced by increasing the clock frequency. Here this figure is 3.7 GHz. And when turbo mode is activated, the processor accelerates to an incredible 4.6 GHz.

Core i5-9600K is the best current processor from Intel today. Then there are models for those who are greedily trying to accumulate as much power as possible for the years to come.

When using an i5-9600K and a good video card, sufficient RAM and other adequate technical characteristics, you shouldn't have any performance issues with modern games.

Intel Core i5-9600K

No. 3 – Intel Core i7-8700K

Price: 23615 rubles

So we moved on to the most powerful Intel line - Core i7. We will begin our consideration with such a model as the Core i7-8700K. There are the same number of cores as in previous models - 6, and the maximum clock speed is the same.

But the i7-8700K has a significantly increased amount of cache memory - 12288 KB. Also, a more powerful graphics core HD Graphics 630 at 1200 MHz was installed here.

12 threads provide a significant power reserve, thanks to which the Intel Core i7-8700K will be relevant for many years to come. The fact that if you have the appropriate video card, all modern games will run even on ultra settings is probably not worth mentioning, this is already clear.

Intel Core i7-8700K

No. 2 – Intel Core i7-9700K

Price: 34299 rubles

The Intel Core i7-9700K processor is based on the architecture code name Coffee Lake-R. It has 8 cores and is created according to the 14 nm technical process standard. The clock frequency of the processor cores is 3.6 GHz, and the cache memory is 12 MB.

Essentially, the Core i7-9700K repeats the previous model, but already contains 8 cores and 16 threads, which further increases the processor's power reserves.

With such a processor, you can not only play, but stream modern games in good quality. There is also an unlocked multiplier and, as a result, the ability to overclock the cores.

The only problem is the very high price, but you have to pay a lot for the power.

Intel Core i7-9700K

No. 1 – Intel Core i9-7960X

Price: 113,030 rubles

Here we come to the first place where the Core i9-7960X is located - this is the best latest generation processor from Intel to date.

It costs three times more than the previous model, but this is more than justified, because there are as many as 16 cores operating at a clock frequency of 2.2 GHz. In turbo mode, it is possible to overclock the frequency to 4.2 GHz. It is supported by a 22 MB cache memory.

If you have a lot of money, you can buy this processor and not worry about your computer not being able to handle anything for many years to come. But if you only need modern games, you can choose something cheaper.

Intel Core i9-7960X

Above are the best processor models from Intel. Among them, you can easily choose an option that will suit your needs and financial capabilities, because all the chips presented here are the best solutions for their price.

Almost every year a new generation enters the market central processing units Intel Xeon E5. Each generation alternates between socket and process technology. There are more and more nuclei, and heat generation is gradually decreasing. But a natural question arises: “What does the new architecture give to the end user?”

To do this, I decided to test the performance of similar processors different generations. I decided to compare models from the mass segment: 8-core processors 2660, 2670, 2640V2, 2650V2, 2630V3 and 2620V4. Testing with such a generational spread is not entirely fair, because Between V2 and V3 there is a different chipset, a new generation of memory with a higher frequency, and most importantly, there are no direct peers in frequency among the models of all 4 generations. But, in any case, this study will help to understand to what extent the performance of new processors has increased in real applications and synthetic tests.

The selected line of processors has many similar parameters: the same number of cores and threads, 20 MB SmartCache, 8 GT/s QPI (except 2640V2) and the number of PCI-E lanes is 40.

To assess the feasibility of testing all processors, I turned to the results of the PassMark tests.

Below is a summary graph of the results:

Since the frequency is significantly different, it is not entirely correct to compare the results. But despite this, conclusions immediately arise:

1. 2660 is equivalent in performance to 2620V4
2. 2670 is superior in performance to 2620V4 (obviously due to frequency)
3. 2640V2 sags, and 2650V2 beats everyone (also due to frequency)

I divided the result by frequency and got a certain performance value at 1 GHz:

Here the results are more interesting and clear:

1. 2660 and 2670 - an unexpected turnaround for me within one generation, 2670 is justified only by the fact that its overall performance is very high
2. 2640V2 and 2650V2 - a very strange low result, which is worse than the 2660
3. 2630V3 and 2620V4 are the only logical growth (apparently due to new architecture...)

After analyzing the result, I decided to weed out some of the uninteresting models that are of no value for further testing:

1. 2640V2 and 2650V2 - an intermediate generation, and not very successful, in my opinion - I’m removing them from the candidates
2. 2630V3 is an excellent result, but it costs unreasonably more than 2620V4, given the similar performance and, moreover, this is the outgoing generation of processors
3. 2620V4 - reasonable price (compared to 2630V3), high performance and, most importantly, this is the only model of the latest generation 8-core processor with Hyper-threading on our list, so we definitely leave it for further tests
4. 2660 and 2670 - an excellent result in comparison with 2620V4. In my opinion, it is the comparison of the first and last (at the moment) generations in the Intel Xeon E5 line that is of particular interest. In addition, we still have sufficient stocks of first-generation processors in our warehouse, so this comparison is very relevant for us.

The cost of servers based on 2660 and 2620V4 processors can differ by almost 2 times, not in favor of the latter, so by comparing their performance and choosing a server on V1 processors, you can significantly reduce the budget for purchasing a new server. But I will tell you about this proposal after the test results.

For testing, 3 stands were assembled:

1. 2 x Xeon E5-2660, 8 x 8Gb DDR3 ECC REG 1333, Intel SSD Enterprise 150Gb
2. 2 x Xeon E5-2670, 8 x 8Gb DDR3 ECC REG 1333, SSD Intel Enterprise 150Gb
3. 2 x Xeon E5-2620V4, 8 x 8Gb DDR4 ECC REG 2133, SSD Intel Enterprise 150Gb

PassMark PerformanceTest 9.0

A more detailed testing report allows us to draw some conclusions:

1. Mathematics, incl. and floating point, mainly depends on frequency. The difference of 100 MHz allowed the 2660 to outpace the 2620V4 in computational operations, encryption and compression (and this despite the significant difference in memory frequency)
2. Physics and calculations using extended instructions are performed better on the new architecture, despite the low frequency
3. And, of course, the test using memory was in favor of V4 processors, since in this case different generations of memory were competing - DDR4 and DDR3.

It was synthetic. Let's see what specialized benchmarks and real applications show.

Archiver 7ZIP

Here the results have something in common with the previous test - a direct link to the processor frequency. It doesn’t matter that slower memory is installed - V1 processors confidently take the lead in frequency.

CINEBENCH R15

The Xeon E5-2670 pulled up the frequency and beat the 2620V4. But the E5-2660, which has a not so visible advantage in frequency, lost to the 4th generation processor. Hence the conclusion - this software uses useful additions new architecture (although perhaps it’s all a matter of memory...), but not so much that it would be a decisive factor.

3DS MAX + V-Ray

The results were similar to CINEBENCH: the Xeon E5-2670 showed the lowest rendering time, and the 2660 could not beat the 2620V4.

1C: SQL/File

When testing a database with file access, the E5-2620V4 processor confidently leads. The table shows the average values ​​of 20 runs of the same test. The difference between the results of each stand in the case of a file database was no more than 2%.

A single-threaded SQL database test showed very strange results. The difference turned out to be insignificant, given the different frequencies of the 2660 and 2670, and the different frequencies of DDR3 and DDR4. There was an attempt to optimize the SQL settings, but the results turned out to be worse than they were, so I decided to test all the stands on basic settings.

The results of the multi-threaded SQL test turned out to be even more strange and contradictory. The maximum speed of 1 thread in MB/s was equivalent to the performance index in the previous single-threaded test.

The next parameter was maximum speed(all streams) - the result was almost identical for all stands. Since the results of different runs fluctuated greatly (+-5%) - sometimes they were at different stands with a significant gap in both directions. The same average multi-threaded SQL test results lead me to 3 thoughts:

1. This situation is caused by an unoptimized SQL configuration
2. The SSD became a system bottleneck and did not allow the processors to overclock
3. There is almost no difference between the frequency of memory and processors for these tasks (which is extremely unlikely)

The result for the “Recommended number of users” parameter also turned out to be inexplicable. The average result of 2660 turned out to be the highest - and this despite the low results of all tests.
I will also be glad to see your comments on this issue.

Conclusions

Of course, I am not saying that there is no difference at all between the processors, I just want to note that for certain applications there is no point in a “planned” transition to a new generation.

If you do not agree with me or you have suggestions for testing, the stands have not yet been dismantled, and I will be happy to test your tasks.

Economic benefit

Below is an approximate calculation.