by Ilya Gavrichenkov
02/22/2011 | 11:42 AM
You should have already heard that Intel has temporarily stopped shipping their chipsets for Sandy Bridge processors because of the recently found error in them. This resulted in much more serious consequences, such as the termination of LGA1155 mainboards sales and even recalls of some systems built on the new CPUs. However, despite these events, we decided not to give up posting articles dedicated to the new processors. First, the discovered error is only affecting the chipset SATA controller and doesn’t concern the actual processors at all. Second, Intel promises to resume shipping their corrected chipsets in about a few weeks from now already, so LGA1155 systems and components should come back to the market. All in all, these problems do not seem to have anything to do with the processors at all, and when things go back to normal again, everything we have to say in our today’s review will be definitely current.
We have every reason to consider Intel a success in the graphics market. Although they have given up chips for discrete graphics accelerators back in the previous century (I hope some of you remember the epochal Intel 740), Intel’s share in the graphics market hasn’t been below 50% for the past few years already. The key to Intel’s success is the smart way of promoting integrated graphics solutions, which are extremely popular and are used in a variety of mobile as well as desktop systems. Moreover, there are no reasons to believe that this share may drop any time soon. On the contrary, when Intel started their production of high-performance Clarkdale and Arrandale processors with the integrated graphics core last year, their positions in the graphics market may improve even more, because now users get Intel graphics with the processor and therefore see no reason to replace it with anything else.
The increasing CPU integration, when processors gradually adopt new functional units, is not the only reason why Intel is so successful in the graphics market. The company is not wasting any time and is gradually improving the performance as well as functionality of their graphics cores. Intel HD Graphics core that was added to Clarkdale and Arrandale processors last year, turned out to be a big step forward. It lifted the integrated graphics performance to a new level, by allowing users to run not only the old 3D games, but also a number of new games. Of course, these are not any of the latest 3D shooters, but mostly games like The Sims and World of Warcraft, which are in fact just as popular.
The recent launch of the Intel Sandy Bridge processors may be regarded as another important milestone on this path. Further processor integration led to the integrated graphics core being placed on the same semiconductor die as the other CPU functional units and therefore direct communication between them. This turned out to be a surprisingly successful solution, so now the integrated graphics core has sped up so significantly, that it may even compete against entry-level discrete graphics accelerators. At the same time, we should also keep in mind that the new graphics core is extremely energy-efficient: the developers made sure that it surpassed inexpensive graphics cards in performance-per-watt aspect. As a result, Sandy Bridge becomes a desirable component for notebooks and contemporary energy-efficient desktops, such as home theater PCs. At least this is what Intel claims.
And what is the real state of things? Our today’s test session will answer this question for you. We are going to study the new generation graphics cores – Intel HD Graphics 2000 and Intel HD Graphics 3000, which are used in mobile and desktop Core processors from the new Sandy Bridge generation.
If we look at Intel’ integrated graphics core from a user prospective, namely as some kind of a “black box” producing 3D images, then we can definitely talk about revolutionary improvements brought by the launch of Sandy Bridge. If we compare Intel’s integrated graphics cores from the current and previous generations, these changes will be noticeable not only in performance aspect, but also in the considerably larger list of games compatible with Sandy Bridge.
However, if we take a pick inside the “black box” of the Sandy Bridge graphics core, we will not find any technological or architectural innovations there. In fact, this is why Intel didn’t eve change the Intel HD Graphics name and simply added a numeric index of 2000 or 3000 to it. The same is true for the architecture: the new graphics core is very similar to the previous generation Ironlake graphics core used in Clarkdale and Arrandale processors.
So, where does this significant performance improvement come from? The answer to this question is lying on the surface: everything has to do with the changes in the CPU internal structure. Integration is not just a catchy word from the promotion banners. Introduction of 32 nm process and packaging the computational cores, graphics core, cache-memory and memory controller within the same semiconductor die allowed engineers to eliminate the reason why integrated graphics couldn’t compete with discrete graphics before – relatively slow work with the memory sub-system. The thing is that integrated graphics cores that had to use part of the system memory for their needs could only exchange data with the memory at a limited speed. Regular DDR2 and DDR3 memory used in contemporary platforms in most cases offers lower bandwidth than special video memory. Besides, the integrated graphics core has to share the bandwidth with the processor computational units as well.
Of course, Sandy Bridge cannot solve this problem completely. The platform built around these processors doesn’t have an option for dedicated high-speed video memory. However, new processors can share their internal resources with the integrated graphics core, which has now become possible specifically due to higher level of integration.
All internal processor units in Sandy Bridge are connected via Ring Bus, and the graphics core is one of them, too. Due to this innovation, the integrated graphics core doesn’t work with the system memory directly, but does it the same way as the computational cores - through the high-speed L3-cache, which is 6 MB or 8 MB big in the new processors. In practical terms, the involvement of L3 cache with the graphics core increases the texturing speed and minimizes the idling of the graphics execution units, while they are waiting for the data to be delivered to them.
Of course, it is not only due to the use of cache-memory, that the integrated graphics in sandy bridge performs the way it does. The execution units in the heart of Intel HD Graphics have also undergone some important improvements. According to the developers, they managed to almost double their bandwidth in the entire number of operations and achieve better parallel performance. The modifications provided the new graphics core with support for OpenGL 3.0, Shader Model 4.1 and DirectX 10.1 standards.
Overall, the internal structure of the Intel integrated graphics core remained the same. Intel HD Graphics contains up to 12 scalar 128-bit execution units. It doesn’t look too impressive especially against the background of the entry-level graphics cards from AMD and Nvidia, with a few dozens of execution units. However, Intel’s units perform better due to internal parallel structure, so don’t be discouraged by the specifications of the Intel graphics core. Just remember that 12 execution units from the graphics core in Clarkdale processors perform at least as good as Radeon HD 4290 core built into AMD processors that has 40 execution units.
In addition to execution units optimization in Sandy Bridge, they also increased the frequency of the integrated graphics core. The previous Intel HD Graphics modification in Clarkdale and Arrandale processors located on a separate 45 nm semiconductor die worked at 900 MHz frequency. The today’s 32 nm production process used for Sandy Bridge manufacturing allowed clocking the graphics core at frequencies beyond 1 GHz, which also contributed to the overall performance of the new Intel HD Graphics core.
However, the term “graphics core frequency” applied to Intel HD Graphics 2000/3000 is a pretty relative thing. The thing is that Turbo Boost technology that changes the CPU clock frequency depending on the current workload, has also spread its effect on the graphics core in Sandy bridge processors. That is why the actual graphics core frequency may change dynamically depending on the power consumption and heat dissipation of the CPU computational units. In other words, the graphics core accelerates when the processor cores are not fully utilized and slows down when CPU power consumption threatens to exceed the limits because of heavy load.
The only thing that remained unchanged since the previous version of the Intel graphics core is its interaction with the image displaying devices. Sandy Bridge has a special independent FDI bus (Flexible Display Interface) that works via DisplayPort protocol. It transfers the image from the GPU inside the processor to the chipset, and from there it is relayed through digital and analogue graphics outs on the mainboard. So, Sandy Bridge chipsets should also feature this FDI bus and video signal routing capabilities in order to support graphics integrated into the new Intel processors. Today this functionality is only offered by the mobile chipsets or the desktop Intel H67 chipset, which also supports HDMI 1.4, unlike chipsets for Clarkdale and Arrandale CPUs.
When we discussed architecture and major features, we spoke about two versions of the new Intel HD Graphics (version 2000 and version 3000) at the same time. In fact, these two modifications have dramatically different performance. And the primary reason for that is Intel decision to differentiate them not only by clock frequency, but also by the number of execution units. This is the major distinguishing feature between Intel HD Graphics 2000 and 3000: the top model has twelve execution units, while the junior model has only six. So, the peak performance of the top model may be at times twice the performance of the junior model.
We totally understand when Intel’s desire to separate the graphics corer targeted for different markets is coming from. Especially, since they intend to use three types of semiconductor dies in their Sandy Bridge processors: a quad-core and dual-core die with a fully-functional graphics core or a dual-core die with limited-functionality graphics. However, the way they position Intel HD Graphics 2000 and 3000 is a little confusing.
The general rule is the following: all mobile Sandy Bridge modifications are equipped with the top model - Intel HD Graphics 3000, while the desktop Sandy Bridge modifications receive a slower Intel HD Graphics 2000 with a few exceptions. These exceptions are Intel K-series processors for overclockers and enthusiasts. Just like their mobile brothers, these CPUs can boast a high-performance Intel HD Graphics 3000 core.
It is quite logical that they would want to use a faster graphics core modification in mobile CPUs. Integrated graphics is needed in notebooks, and even if they are equipped with a stand-alone graphics card the manufacturers retain the option to switch to more energy-efficient graphics core integrated into the processor, which will now be fit to handle more situations. However, it is at least strange to deprive almost all desktop users of the Intel HD Graphics 3000 accelerator. High-performance integrated graphics may come in very handy in desktops, too. But according to Intel, even the slow HD Graphics 2000 will do the job just fine, even though the faster modification of this graphics core could easily replace an entry-level graphics card. And it is even more surprising to see that Intel integrated their fastest HD Graphics 3000 modification into the Core i5-2500K and Core i7-2600K processors that do not need it at all. These CPUs are positioned for computer enthusiasts, who will never use integrated graphics merely because the Intel H67 chipset supporting it doesn’t offer any overclocking-friendly options at all.
The reasons behind this decision may be coming from the fact that they believe their HD Graphics 3000 is fast enough only for notebooks, because they tend to have lower screen resolution than desktop monitors. And since they do not think the performance of any integrated graphics core is high enough for comfortable work in desktop 3D applications, they modestly offer a junior graphics core in desktop CPUs. That could definitely be the case. But why would they put a high-performance graphics core into their Core i5-2500K and Core i7-2600K processors, while most users would totally understand if there were none? I can’t answer that.
Besides using graphics cores with different number of execution units in different Sandy bridge processors, Intel also resorted to an old time-tested trick: clock speed differentiation. Different processor models may have graphics cores working at different clock frequencies. However, now Intel HD Graphics frequency is determined strictly by the thermal envelope, instead of being a justification for market positioning.
The table below contains the general info about the Intel HD Graphics modifications integrated into second generation Core processors. Note that we have two frequencies for each graphics core: turbo-frequency at which the graphics core works in most cases, and nominal frequency down to which the graphics core may slow when the computational CPU cores are heavily loaded with work.
So far we have spoken of Sandy Bridge graphics core as the next evolutionary step from Intel HD Graphics used in Clarkdale and Arrandale processors. However, it does have something unique about it too and that is Quick Sync technology providing hardware acceleration of HD video encoding and decoding.
You may think there is nothing surprising about it. All GPUs from AMD and Nvidia have long been able to encode and decode video using CUDA or Stream/APP. Moreover, previous generation Intel graphics integrated into Clarkdale and Arrandale processors is also capable of accelerating video playback on the hardware level. However, Intel engineers used a completely new approach to this matter in Sandy Bridge. The thing is that all GPUs that have been out there so far use their default execution units, namely shader processors, to work with video. Quick Sync technology works differently: it implies that there will be specific execution units involved into the encoding/decoding process. In other words, Sandy Bridge graphics core has a dedicated video codec and video decoder besides the traditional execution units.
Of course, the use of utilitarian hardware resources instead of a software/hardware combination with CUDA or Stream is not a universal solution, which also increases the die size. But according to Intel, the advantages of this solution are overpowering. Firstly, Quick Sync delivers higher performance, and secondly, special hardware turns out much more energy-efficient.
So, Quick Sync consists of two components. The first one is hardware decoder used to accelerate the playback of video content in popular MPEG-2, VC-1 and AVC formats. This part of Sandy Bridge graphics core can take over the entire decoding process including motion compensation and loop-filtering. Most importantly, this is a multi-threaded decoder, i.e. it can decode video in several parallel threads at the same time supporting picture-in-picture mode, stereo 3D Blu-ray or MVC.
The second part of Quick Sync is a hardware codec performing the opposite operations. Unlike the decoder, the codec also utilizes the traditional execution units of the graphics core, although most of the encoding is still done by special logics. The codec supports today’s most popular AVC format.
As a result, Quick Sync technology as a whole allows speeding up video transcoding by decoding video stream in one format and encoding it right away into another. This is a very important and timely feature of the new Sandy Bridge processors. Video content transcoding operations are very popular even among home users, as video content hosting services as well as increasing popularity of high-performance mobile devices capable of playing multimedia content become more and more widespread.
As we know from our previous processor test sessions, video transcoding involving traditional processor capacities is pretty much the most resource-consuming operation, which takes a lot of time and uses a lot of power. With Quick Sync technology we can not only speed up this process, but also free processor cores for simultaneous execution of other tasks.
Of course, Quick Sync technology must be supported on the software level. But Intel didn’t forget about it and today there are a lot of popular utilities for video transcoding that can utilize effectively special units in the new Sandy Bridge processors. And the examples are right in front of us: new versions of ArcSoft MediaConverter, Corel DVD Factory, CyberLink MediaEspresso, Movavi Video Converter, Roxio Creator and other applications already support new media functionality.
Once Intel integrated new codec and decoder into their graphics core and ensured widespread support from software developers, they managed to outplay AMD and Nvidia who offer slower video transcoding involving their shader processors. However, we do not know if isolation the codec and decoder into individual hardware units will ever grow into a new trend. Maybe Quick Sync technology will simply encourage other GPU developers to optimize their transcoding algorithms using CUDA and Stream/APP.
However, Intel slightly spoiled the good impression from Quick Sync. It was a big mistake on the developers’ part to place encoding and decoding units into the graphics core. Video transcoding is something you would do independent of the type of graphics, integrated or discrete, in your system. But unfortunately, you won’t be able to use Quick Sync resources in Intel P67 based systems. Mainboards with this chipset will disable the integrated graphics core and thus you lose access to this promising technology. So, you will only be able to take advantage of hardware acceleration in the Intel graphics processor in those systems where the integrated Sandy bridge graphics core is working.
Intel H67 chipset designed for use with graphics core in the Sandy Bridge CPUs doesn’t support overclocking. It provides no access to the multiplier for the CPU as well as memory frequency. However, Intel H67 allows adjusting the graphics core frequency. Mainboards based on this chipset usually have an option showing how high the graphics core can be overclocked in Turbo mode.
And this option is not just for show. Intel limits the graphics core frequency in coordination with the TDP of the corresponding processor models, which means that it is quite possible to overclock with proper cooling.
We performed overclocking experiments with several different Core i5 processor samples and found that with a slight increase of the graphics core voltage we could easily push its frequency to 1.5 GHz. In other words, the GPU inside sandy Bridge processors doesn’t have that big of an overclocking potential, but we can still count on at least 30-40% graphics performance improvement during overclocking.
However, this is not the case for Core i7 CPU models. The frequency of Intel HD Graphics 3000 core inside them is almost at the maximum already.
During our today’s test session e decided to compare the performance of the new Intel HD Graphics 2000 and Intel HD Graphics 3000 accelerators integrated in Sandy Bridge processors against the competitor integrated GPUs and entry-level graphics cards. There are currently two integrated graphics platforms for desktops in the market that could be considered worthy competitors to Sandy Bridge: Socket AM3 with AMD890GX chipset and LGA1156 with Clarkdale processors and Intel H57/H55 chipset (we do not take into account their numerous modifications here). These will be the today’s competitors to the systems based around second-generation Core i5 processors with Intel HD Graphics 2000 and Intel HD Graphics 3000 cores. Moreover, we also included inexpensive discrete graphics accelerators from AMD: Radeon HD 5450 and Radeon HD 5570.
Since we couldn’t compare the integrated graphics ores in CPUs of the same computational capacity, we used Socket AM3 and LGA1156 processors with the today’s highest clock rate, which could at least somehow compete against Sandy Bridge. Note that we had to work with two Clarkdale models – Core i5-680 and Core i5-661. The former has the highest clock frequency, while the latter features higher –performance graphics core. In the AMD platform we used a quad-core Phenom II X4 975. As for Sandy Bridge, we took Core i5-2400 to test Intel HD Graphics 2000 and Core i5-2500K to test Intel HD Graphics 3000. As for the discrete graphics cards, we tested them in the same LGA1155 system with the Intel Core-i5 2500K processor.
As a result, we ended up putting together the following testbeds:
We tested Intel HD Graphics 2000 and Intel HD Graphics 3000 graphics cores in two modes: at nominal (for most desktop processors) frequency of 1100 MHz and at an overclocked frequency of 1500 MHz.
We would like to start discussing the test results with the general performance benchmarks. Although, these performance numbers do not have direct connection to the primary topic of our today’s discussion, they may be very useful for further analysis of the obtained results. You can refer to our previous article for detailed performance tests of the new Sandy Bridge processors and their competitors. Here we used only PCMark Vantage test that shows the performance during execution of some typical algorithms.
The results charts show serious diversity. No matter how you put it, but LGA1155 platform boasts significantly higher computational performance than all other solutions. So, if you are looking to build an integrated system with maximum performance in computational tasks, then your choice is obvious.
Since all versions of Intel HD Graphics support DirectX 10, but not DirectX 11, we used specifically the Vantage benchmark to estimate the performance of the ne graphics core modifications. And obtained results showed that even Intel HD Graphics 2000 core with only six execution units is considerably faster than the graphics core in Clarkdale processors, even though the latter has twice as many execution units inside. However, only Intel HD Graphics 3000 could compete against discrete graphics accelerators. Although even when we overclocked it to 1500 MHz, it couldn’t compete against Radeon HD 5570, so Sandy Bridge processors can in fact only compete against $50-$60 graphics cards.
Intel HD Graphics 2000 offers pretty comfortable gaming experience in 1024x768 resolution, and the faster Intel HD Graphics 3000 can also allow you to play in HD. Of course, you can only use the lowest image quality settings, but only in this case integrated graphics can deliver more or less acceptable results.
Note that overclocking of the Sandy Bridge graphics core provides a substantial performance boost, but even with this performance gain Intel HD Graphics 3000 can’t catch up with Radeon HF 5570. However, if we compare the new Intel integrated GPU against integrated competitor solutions, then even the limited Intel HD Graphics 200 model will leave all of them far behind.
I have to say that Intel HD Graphics 3000 does just fine as a decent entry-level 3D graphics accelerator. You can get a decent fps rate even in 1680x1050 resolution by playing with the image quality settings. Too bad that this GPU modification is not available in most desktop CPUs. The cut-down modification, Intel HD Graphics 2000, works considerably slower, and even overclocking doesn’t allow it to reach the performance of the top Sandy Bridge GPU.
Metro 2033 is a very demanding game. No wonder that Intel HD Graphics 3000 can barely handle it. However, you will get much better experience if you use a discrete graphics accelerator. But obviously not any graphics accelerator will do (Radeon HD 5450 is pretty much as fast as Intel HD Graphics 2000) – you will need a pretty fast one, at least as fast as Radeon HD 5570.
Contemporary 3D shooters are not designed to be run on integrated graphics. The fps rates in Mafia II are even lower than in Metro 2033, so the only way you can achieve acceptable performance is if you overclock Intel HD Graphics 3000 to its maximum. To be fair we have to say that the performance of other integrated graphics cores from AMD and Intel less than half of Intel HD Graphics 3000.
The results in Lost Planet 2 are particularly interesting because they show very well show the differences in performance of two Sandy Bridge graphics core versions with 6 and 12 execution units. But even a non-overclocked Intel HD Graphics 2000 is significantly faster than any of the previous-generation graphics cores out there.
Notebooks based on Sandy bridge processors that use their integrated graphics core will most likely deliver acceptable gaming performance. As we see, Intel HD Graphics 3000 graphics core can provide decent performance in 1680x1050 resolution in this popular racing game, just like in many other games. The only thing you will need to do is to set the image quality to the minimum.
H.A.W.X. 2 proves to be a not very demanding game. Even the owners of previous-generation integrated graphics will be able to enjoy it. Intel HD Graphics 3000 and Intel HD Graphics 2000 compete successfully against discrete graphics cards in this game.
Contemporary strategies rarely require high-performance graphics cards. And this is exactly the case here. If we do not set the image quality settings too high, then Civilization V will run perfectly fine on both Intel HD Graphics modifications in Sandy Bridge and will even allow using HD resolutions.
Starcraft II is not a typical application: this is where new Intel integrated graphics loses drastically even to the Radeon HD 5450. However, it doesn’t mean that the integrated graphics in Sandy Bridge is not powerful enough for this game. if you set low image quality Intel HD Graphics 3000 will be fast enough in 1680x1050, and Intel HD Graphics 2000 will deliver acceptable fps rate in simpler modes.
Sandy Bridge graphics core has absolutely no problems with HD video playback. The decoder built into the graphics core copes perfectly fine with all popular formats, leaving the computational part of the CPU almost without work.
The only possible issue could be connected with the support of the new Sandy Bridge decoder by software players. For example, two weeks ago, during Core i5-2400S tests we discovered that a popular free Media Player Classic Home Cinema had some issues with the new processors: we observed certain artifacts during DXVA rendering. Therefore, at that time we switched to a commercial Cyberlink PowerDVD10, which was better optimized for Intel HD Graphics 2000/3000 and worked impeccably. Again we had to resort to a commercial product, because even the recently released Media Player Classic Home Cinema version 188.8.131.5227 still has certain issues with Sandy Bridge CPUs.
This is what the hardware acceleration of video playback in a Core i5-2400 based system with Intel HD Graphics 2000 graphics core looks like:
H.264, firstname.lastname@example.org fps, 23.7Mbps
The CPU utilization during HD video playback in H.264 format doesn’t exceed 10-15%, while the CPU itself remains in power-saving mode, i.e. works at the clock speed lowered to 1.6 GHz.
VC-1 video playback in Cyberlink PowerDVD10 (version 10.0.2429.51) produces even more impressive results.
VC-1, email@example.com fps, 16.5Mbps
In this case the CPU utilization is even less than 10%. However, we noticed one peculiarity during our tests: the container type for the decoded VC-1 video stream matters a lot. For example, we couldn’t get hardware acceleration for VC-1 video in MKV container to work, while the same content in MPEG-TS container or played directly from a Blu-ray disc was played back perfectly fine using our hardware decoder.
So, we can conclude that the hardware video decoder in Sandy bridge processors does its job just fine, so these CPUs will be perfectly fit for a media center PC even without an add-on graphics card. Especially, since the computational capacity of the Sandy Bridge processors is sufficient for it to playback any video content, even the most exotic one, without any help from the hardware decoder.
Hardware acceleration during video playback is one half of Quick Sync technology, because it only uses the decoder and doesn’t involve the codec inside the Sandy Bridge graphics core. Only video transcoding reveals all the potential of this technology, as the content is first decoded and then encoded into new format. Some popular video transcoding utilities can already offer video transcoding using special units in Quick Sync instead of the processor capacities. Take, for instance, the latest Arcsoft Media Converter or Cyberlink Media Espresso.
The distinguishing feature of video transcoding in applications using Quick Sync is the low CPU utilization that stays way below 100%. And that is one of the most computationally heavy tasks out there!
For example, the screenshot above shows that transcoding a 1080p H.264 movie into lower resolution and lower bitrate for viewing on an iPhone 4 uses only 15-20% of the CPU. But the most impressive thing about it is that the CPU in this case doesn’t even switch from the power-saving mode and works at a lowered frequency of 1.6 GHz.
The video quality doesn’t suffer at all in this case. This is what positively distinguishes Quick Sync from transcoding via CUDA and Stream/APP. When we compared the quality of the video transcoded using Quick Sync and the video transcoded using traditional processor resources, we didn’t discover any serious artifacts produced by the Sandy Bridge graphics core.
Intel Quick Sync
Moreover, we got the feeling that the video transcoded with Quick Sync turned out even better. Note that the size of the resulting files differed by no more than 2%, and the actual bitrate and resolution were in fact identical. In other words, Quick Sync does work very well.
Transcoding speed also turned out high enough. The use of Quick Sync technology not only offloads the CPU computational cores, but also delivers the final result faster. We performed several tests transcoding a short 1080p H.264 movie fragment into an iPhone 4 compatible format (H.264, 1280x720, 4 Mbps). The diagrams below show how much time it takes when we use processor’s computational resources, when we activate Quick Sync technology, and when we transcode using CUDA and Stream/APP using external AMD Radeon HD 5570 and Nvidia GeForce GT 430 graphics cards. We used the same Core i5-2400 based testbed for all tests. We used two different applications to measure the transcoding speed: Arcsoft Media Converter 184.108.40.206 and Cyberlink Media Espresso 6.5.1229.
Intel has every reason to be proud of their Quick Sync technology. The decision to make the decoder and codec into individual units turned out a great performance-improving solution. As you can see from the diagrams, this technology speeds transcoding significantly compared with the situation when it is performed by processor computational cores. We failed to achieve the same high result with the discrete entry-level graphics cards from AMD and Nvidia, either.
Judging by the performance, the graphics core in Sandy bridge processors could actually replace the entry-level graphics cards easily. Of course, in this case we should also benefit in terms of power consumption. We decided to perform our traditional power consumption tests in order to estimate our potential gain from this replacement.
The graphs below show the full power draw of the computer (without the monitor) measured after the power supply. It is the total of the power consumption of all the system components. The PSU's efficiency is not taken into account. The CPUs are loaded by running the 64-bit LinX 0.6.4 utility. Graphics cores were loaded using FurMark 1.8.2 utility. Moreover, we enabled C1E and Enhanced Intel SpeedStep power-saving technologies to ensure that computer power draw in idle mode was measured correctly.
We have already stressed several times before that Sandy Bridge processors are very energy-efficient in idle mode. The use of the integrated graphics core doesn’t take away this advantage. Systems built on second-generation Core processors with the integrated graphics core consume 5-10 W of power less in idle mode than any other platforms.
LGA1155 systems with integrated graphics also consume less power in processor-heavy tasks. This is quite logical, since Sandy bridge computational cores currently offer the best performance-per-watt ratio. And the graphics core built into these processors is more energy-efficient than external graphics accelerators and other integrated cores at least due to the fact that it is manufactured using the latest 32 nm process.
Heavy graphics load shows very clearly that Sandy Bridge graphics is very energy-efficient. Its performance is close to that of entry-level graphics accelerators, while the power consumption is 10-15 W lower. By the way, note that Intel HD Graphics 2000 and Intel HD Graphics 3000 modifications (the latter has twice as many execution units as the former) differ by only 6 W in power consumption, and the 35% overclocking of the graphics core frequency has even less effect on the power consumption.
Complex work load applied to the computational and graphics resources of the system lets Core i5-2500K compete on equal terms with the platform utilizing Core i5-2500K and its integrated graphics core. However, since the Intel HD Graphics 3000 core built into this processor performs better than the above mentioned graphics card, there is nothing remarkable here. There is another interesting thing about it: the peak power consumption of contemporary integrated platforms using latest generation high-performance processors doesn’t exceed 90 W. so, even the regular (non-energy-efficient) Core i5 modifications can easily be part of a compact home system or HTPC.
Another type of complex load is HD video playback. Due to a special hardware video decoder in the new Sandy Bridge processors, there is barely any load on the computational CPU cores. As a result, the power consumption of Sandy Bridge systems with any graphics core modification is only 6 W higher during video playback than the power consumption in idle mode. As a result, HD video playback in systems like that requires even less energy than some platforms would need for idle mode alone.
It would be hard not to give Intel’s new generation graphics a super high rating. Intel HD Graphics 2000 and Intel HD Graphics 3000 graphics accelerators built into Sandy Bridge processors are in fact a new standard of integrated graphics. By combining processor cores with the graphics core within the same processor die, using a high-speed ring bus to connect them and implementing special optimization into the execution units of the new Intel GPU, they turned the new graphics cores not just the fastest solutions in their category. Moreover, now we can compare them against contemporary discrete graphics accelerators. This is primarily true for the top Sandy bridge graphics core called Intel HD Graphics 3000. It is considerably faster than AMD Radeon HD 5450, for instance.
In other words, Intel HD Graphics 3000 may become a serious threat to $50-$60 graphics cards. In this case, discrete graphics cards have only one important advantage over the integrated Intel graphics core: they support DirectX 11, which is missing in Intel HD Graphics. However, this functionality is hardly necessary for the level of performance provided by these solutions, anyway.
The junior modification, Intel HD Graphics 2000, is considerably slower. However, it still looks very nice against the background of other integrated solutions. I doubt that anything could threaten the leadership of Intel HD Graphics 2000 and Intel HD Graphics 3000 in their segment, at last not until AMD Llano processors come out.
Intel’s Quick Sync technology that appeared in the new Intel cores has also proven extremely successful. Intel offers a principally new approach to HD video encoding and decoding acceleration by introducing special hardware units for that purpose. By making the processor die a little bit bigger Intel managed to achieve impressively high performance and exceptional HD video encoding and decoding quality. And looks like Intel Quick Sync technology transcodes video way better than Nvidia CUDA and AMD Stream/APP.
Unfortunately, Intel’s strategy in respect to the marketing of the new graphics cores has slightly spoilt the great impression left by their features and performance. The top graphics core modification will only be available to the owners of mobile computers, and if you want to have it in your desktop, you will have to go for an expensive overclocker CPU. So, in fact, it would be fair to say that Intel HD Graphics 3000 is targeted for mobile applications, while the desktop users will have a slower Intel HD Graphics 2000 modification only.
On the one hand, it means that notebooks based on second-generation Core processors receive a very fast graphics core by default. So the new mobile computers that do not have any discrete AMD or Nvidia graphics cards will now be also deliver acceptable performance in the majority of 3D apps.
On the other hand, things are a little different in the desktop segment. Intel HD Graphics graphics cores cannot oust the discrete cards from this segment yet. The integrated graphics core in most desktop Sandy bridge processors may be a good choice for a multimedia PC, especially due to superb Quick Sync technology, but its 3D performance will still be insufficient to satisfy even the most undemanding users.