Since AMD's Ryzen 7 3000 family preparation was very hectic, to say the least, it took a lot of nooks to get the Navi and Ryzen exams completed in time. There was also some difficulty in ensuring cross-compatibility of the X470 and X570 motherboards was properly maintained, and not much time wasted watching. For this reason, I chose to test the Ryzen 7 2700X on an older X470 motherboard and the Ryzen 7 3700X and Ryzen 9 3900X on the new X570 motherboard. In our review, I found that AMD had relatively high idle power and that we did not know why.
We have done additional research on this and can now clarify part of the problem: X570 motherboards use a lot more energy than their X470 counterparts. If you put a Ryzen 7 3000 processor in an X470 chipset, suddenly a considerable energy efficiency appears. We changed the power graphics of our initial test and tested an additional X570 motherboard.
The MSI Godlike X570 was our original test bench, but we added a comparison with the old X470 MSI Gaming M7 AC. Although the Godlike X570 does give some of its energy to features such as OLED and integrated LED bling, this does not seem to be enough to match the level of difference. The other X570 motherboard we tested, the Asus Crosshair VIII Hero (Wi-Fi) also has a different power – and lower – but not as much as the OG X470.
We have not yet tested the 3900X, but there is no reason why these results do not apply as well. Here are our adjusted / expanded results.
The passage of the Godlike MSI X570 to the X470 Gaming M7 AC reduces power consumption by 38W under Prime95. Idling power goes from 67W to 52W, a substantial reduction. Suddenly, the AMD Ryzen 7 3700X processor at 7 nm is now only slightly behind the Intel processors, instead of being separated from them by a sinkhole.
The Asus X570 board has lower overall power consumption than the MSI, but not enough to reduce the gap between the two solutions.
Compared to the Ryzen 7 2700X on an even chipset, the gap between them has become a canyon. At full load in the old Prime95 version, the 3700X of AMD consumes about 70 W less.
The new Prime95 shows to play reduction to power on X470 compared to what we measured for X570 in this test. We confirmed the result, but we have no explanation. The two X570 motherboards maintain relatively constant power between the two variants of Prime95, but the X470 motherboard drops 20W. The Godlike MSI uses 15-16W more than the Asus X570 Crosshair VIII Hero.
The gap between the Ryzen 7 2700X and the Ryzen 7 3700X expands here to reach nearly 90W. These reductions give the Ryzen 7 3700X better energy efficiency than the Core i9-9900K, even after the Intel processor has reduced its frequency.
Finally, Cinebench R20. Here, our lowest power line is not standby power, it is a single-threaded power supply during the single-core test. There is a difference in power of 17 W between the cores and if AMD still does not match the relatively low power of the Core i7-8086K, the 80 W is a much better result.
Again, however, the X570 continues to draw considerably more power than before. The MSI card is one of the main offenders, with 174W, but even the Asus card still draws between 160 and 165W, against 130W for the AMD Ryzen 7 3700X. Again, even compared to the Intel Core i9-9900K, the Ryzen 7 3700X is a great choice.
The X570 chipset is extremely power hungry at the moment. Some of these problems can be corrected in UEFI updates. It would also be interesting to have the option to disable all LEDs or built-in OLED panels (such as on the Godlike X570) to confirm that they have no significant impact on the power of the problem.
AMD has always said that the X570 chipset would cover the X470 / B450 rather than replace it, but the company did not really say what energy efficiency it would give up by focusing on these chipsets as launchers. The Ryzen 7 3700X consumes far less energy than it seems at first when it is associated with a less energy hungry chipset. Running PCIe 4.0 for internal bandwidth, LEDs included, built-in fan – even with all of these factors in mind, there is still a huge gap between the two platforms.
In the meantime, when you compare the Ryzen 7 2700X to the 3700X of the same motherboard, the degree of improvement is incredible. From apples to apples, with the motherboard and with PCIe 4.0 support as the only difference, the Ryzen 7 3700X consumes 70 to 90 W less power than the Ryzen 7 2700X.
Update (7/9/2019): The original title of this story, "AMD's X570 chipset consumes so much power, it's a processor comparison," implied that the power differential was entirely a function of the X570 chipset itself. It was an unintentional inaccuracy. This is not the X570 chipset only that pulls the extra power, that's all the motherboard. The current X570 chipset only consumes ~ 7W at load. (If the X570 chipset actually called 30 to 50 W, the size of the heatsink and fan would be much larger than it would be).
The question of why these motherboards consume so much more extra power and why there is so much divergence from one board to the other is still in the study. These are all early motherboards and future updates of the UEFI could solve or significantly improve the situation.
As the launch date of AMD's next Zen 2 architecture is fast approaching, the company has raised the curtain and provided insight into the capabilities and improvements of its new model. These new chips include a number of enhancements and benefits to generate both higher cycle instructions (CPIs) and better overall power.
Let's start with some basics. The Ryzen 3000 family is powered by AMD's Zen 2 architecture except for its APU. APUs actually have a generation behind them – the 2000 Series APUs were actually built on the first generation of Ryzen, and the APUs of the 3000 series are based on the second generation Ryzen. The architectural improvements and other features we are going to talk about today do not apply to the Ryzen 3 3200G or the Ryzen 3 3400G.
During his Zen 2 presentation, AMD Corporate Fellow Mike Clark said his 7nm transition was in fact more successful than originally planned.
Some of you may remember rumors that AMD would use Ryzen 3000 processors with clocks much higher than the previous ones. According to AMD engineers, the company did not necessarily expect Zen 2 to reach higher frequencies. at all. This is the intrinsic problem of shrinkage of modern UC nodes. Smaller process requirements mean lower voltages, and lower voltages can negatively impact the absolute operating frequency. In this case, however, the 7 nm TSMC node and AMD's own engineering were able to create components that could reach slightly higher frequencies than the 12/14 nm chips.
It should be kept in mind that AMD did not expect improvements in the clock frequency of 7 nm, which should be taken into account when doing the same. Evaluation of the accuracy of rumors about massive clock jumps in the future.
An important change coming with Zen 2 has nothing to do with the current processor. AMD informed us during the event that new scheduler changes were being incorporated into Windows 10 Scheduler from Windows 10 1903 (May 2019 update). There are two new features: topology recognition and faster clocking of the clock. A faster acceleration of the clock reduces the time required by the processor to change state, thus improving performance and theoretically, inactive power by allowing the processor to move more quickly to different states. lower clock. Knowledge of the topology should help to keep local CCX data relevant and complete a CCX before loading another one.
These gains – a performance of + 15% in 1080p in Rocket League and a 6% improvement in the launch of PCMark 10 applications – are only result from the update of the Windows 10 scheduler and are separate from any additional gain resulting from improvements to the Zen 2 architecture. Taking advantage of these enhancements requires both an updated chipset driver and the upgrade. day Windows 10 1903.
This slide represents the microarchitectural synthesis of AMD. The chip integrates with the new TAGE branch predictor in addition to the BP perceptron it used in the past. The micro-op cache has been increased to 4K instructions, with double the total of L3 on board. (AMD is now referring to its L2 and L3 combinations as "AMD GameCache".) A new Address Generating Unit (AGU) is now associated with the entire side of the core, with full support of the comma floating 256-bit via AVX2.
The slideshow below presents our in-depth analysis of the specific architectural improvements of the third-generation Ryzen processor. Each slide can be clicked to open it in a new window.
According to AMD, these improvements give them an edge over Intel, both in terms of performance per watt and absolute wall power.
Cinebench is not the ideal solution for measuring energy consumption, but it is not a bad test either. The 3700X – which, in all fairness, is probably closer to the ideal position for architecture – is assumed to be 56% more efficient than the Core i7-9700K, while it only consumes 86 % of the power in absolute value.
Energy efficiency gains over the 2700X are even greater. AMD claims that the 3700X is 1.75 times more efficient in performance / watts than the 2700X, while consuming 70% power.
Although we obviously can not decide to launch before having material to test, AMD offers an aggressive and exciting product family. The TDPs have dropped dramatically. The CPI would have risen 1.15 times. Clock speeds have been increased. Planner improvements and doubled floating point capacity should provide their own robust improvements beyond this figure of 1.15. The width of the Infinity Fabric bus has been doubled to allow the use of PCIe Gen 4 bandwidth and a new memory divider at the DDR4-3733 level can reduce IF clocks without compromising the time of day. DRAM scale.
If you're a fan of AMD APUs, 7 nm also has exciting long-term implications for them. Although we do not know when we will see these parts, the company has clearly aggressively targeted a decline in power in all areas. This will clearly pay off when refreshing the APU family at 7 nm. According to one of our 7nm launch theories, AMD would focus on energy efficiency at least on some parts, and we see it perfectly, with a higher-performance 8-core processor in a 65W TDP and a 16-core processor in at 105W TDP.
Intel has not only detailed its new Ice Lake architecture at Computex this week, it has also announced a new processor: the Core i9-9900KS. This chip is an eight-core spin on the 9900K, but with an even higher core all-core clock of 5 GHz under load. This represents an increase of 300 MHz over the standard Core i9-9900K, with a higher base clock of 400 MHz.
We mapped how the CPU will fall in the global stack of Intel, with AMD comparative parts listed. We have also included the Ryzen 7 2700X, for reference. Prices marked with a ~ are taken from retail prices from 5/28. The asterisk near the TDP of the 9900K refers to the fact that it has been shown that this chip was only a 95 W processor if it was limited to an operating clock of 3 , 6 GHz. At higher speeds, the power consumption of the processor increases dramatically. At full load, the existing version of the Core i9-9900K can display power from 160W to 180W. The increase in clock speed to 5GHz all-heart, assuming that the chip keeps this frequency at all times (as it should be), will probably exceed the threshold of 200 W – more, if the voltage taps required to reach this frequency are significant.
Critics who viewed the Intel Core i9-9900KS processor demo said it was using high-end, but ultimately standard cooling components, with a 240-millimeter ROG Ryujin 240 chiller. It's a high-end system, but not crazy. Those who fear a repetition of the 28-core processor with a 1.2-kW cooler attached to the table do not have to worry.
This processor will be available at retail, with fully functional integrated graphics, at an undisclosed price. The overall positioning seems clear. Intel will push the 9900K to the extreme to give it a better chance of beating the Ryzen 7 3800X in a heart-to-heart comparison. The chances of matching the 12 AMD cores are slim, unless this chip turns out to have a really terrible scaling because of the bandwidth pressure of the memory (and for to be clear, this is not the expected result). Although we are not surprised to see that the 12-core Ryzen 7 evolves less efficiently than its Threadripper counterpart, moving from 8 to 12 cores increases overall resources by 1.5 times. Even with a low scale, this can still lead to a significant improvement in performance.
So, why does Intel have to choose this path? A little marketing and a little strategy. All-heart 5GHz technology is a pretty powerful marketing claim, and not all applications go beyond eight cores. In fact, the general rule you can count on is that the higher the number, the worse the scale is (outside the so-called "embarrassingly parallel" workloads that scale exceptionally well). ). Many multithreaded applications have also exceeded 4 or 8 threads, because until recently, 16-core processors remained the only domain of workstations or high-end servers.
The price is an interesting question. To date, Intel has shown little interest in lowering its price per core below certain thresholds. Although the company's prices have improved significantly since the initial launch of Ryzen (an Intel eight-core chip is now literally less than half the price of 2016), the company has so far allowed AMD to claim the crown price / performance, is fighting for the absolute performance crown. We will have to see how the 3800X and 9900KS models compare in face – to – face competitions before knowing which of them will lead overall.
AMD has finally unveiled its Ryzen 3000 family, with specific heart counts and boost frequencies. The company's 7-nm products have been the subject of much speculation over the past six months as enthusiasts eagerly await the company's next generation of products.
The new AMD product goes from a 12-core option at the top for $ 499 to a six-core processor for $ 199. All processors now support PCIe 4.0 and TDPs have improved significantly. The previous 12-core AMD processor, the Threadripper 2920X, is designed for a 180W TDP. The new 12-core processor is rated for 105W TDP. The 4.3 GHz boosting frequency of the 2920X has been improved to 4.6 GHz, again, it is a maximum acceleration, not a full optimization. The cost of the processor is also significantly better when switching to the Ryzen platform instead of using Threadripper.
The 8-core Ryzen 7 3800X improves the base and amplifies the clocks of the 200 MHz Ryzen 7 2700X, but does not reduce the nominal TDP of the processor. We expected, however, that the current TDP as measured on the wall would decrease, as the 12 cores are sliding in the same maximum rated capacity of 105W and the 3700X is actually a 65W processor.
The planned product stack was almost completely wrong, although the details vary depending on the product. Prices were incorrect – far too low – in all cases. There is not $ 100 six hearts. There is not $ 500 16-core. The product name was incorrect. The Ryzen 9 3900X almost corresponds to the supposed "Ryzen 7 3700", except that the expected price is too low by 40% and the TDP is off by 10W. It is, however, the closest escape from the accuracy of the initial leak, and to be as precocious as it was, the accuracy is excellent. But that's one part out of 10.
The Ryzen 7 3800X is the closest to the supposed "Ryzen 5 3600X", at 3.9 / 4.5 instead of 4.0 / 4.8, and costs $ 400 instead of $ 229. There is still no known APU on the market in 2019 in the planned configurations; AMD you stated it will not use the Matisse design to build an APU. We have no date on APU introductions at 7 nm, but we would not expect them until the end of 2019 or the beginning of 2020 at the earliest.
These rumors were still too aggressive to be accurate given the current state of node narrowing. The price is always the latest That a company is installing, and AMD certainly had prices on the Ryzen 3000 family from last December. AMD was never going to empty its profit margins or disrupt all of its products as planned. The company should have organized a discount sale on its hardware at 14 nm to clean up its inventory and market 7 nm.
The reason AMD was able to do such a reset of its own products with the initial launch of Ryzen is that Ryzen packaged a 1.52-fold improvement in CPI over Piledriver. The planned improvement of 1.15 times for the third generation of Ryzen is not large enough to allow the same radical reshuffling of prices and product segments. AMD told Wall Street that its launch in 7nm would improve margins and not damage them. As some have suggested, the chiplets were not a magic solution to these fundamental problems of scaling and power.
The goal here is not to say that you are right. The goal is to inform you about how these product decisions are made in the first place so that you can assess the accuracy of the rumors. for yourself. No one, including me, is right all the time. Everyone, including myself, believed rumors that were false or did not believe the rumors that were true. But there were good objective reasons not to believe that AMD would dump its pricing by heart and improve the overall efficiency of the processor and reduce the TDP, and greatly increase the clock speed and double the maximum number of processor cores available on its desktop platform.
AMD has clearly decided to target energy efficiency on a raw clock with its 7 nm core, which can be the result of a sharp increase in power when the clock exceeds a certain point. This is not clear at the moment, but the Ryzen 7 3700X is a remarkable chip, at 65W. At 399 and 329 dollars, the latest AMD eight-core chips have again sharply reduced Intel's positioning and pricing.
AMD also claims to be competitive on IPC, with a performance improvement of 15% over Ryzen 7. We expect a performance increase of 15 to 20% with a mix of clock and clock improvements. 39, efficiency, but we will have to wait for material. to see if the claimed earnings are checking. There are still some facts that we have not yet understood, such as the expected improvements in memory support. Rumor has it that the DDR4-3200 is officially supported, with potentially improved support for higher frequencies like DDR4-3600. Support for PCIe 4.0 is also welcome, although we expect this to be particularly useful with SSD speeds – single-GPU configurations, historically, do not gain much in performance (if any) with the bandwidth enhancements provided by PCIe at the time of its introduction.
Overall, this is a solid improvement from Red Team (I bow to the inevitable weight of history until the name "Team Green" is abandoned, since AMD has adopted Crimson under his identity). The introduction of a 12-core high-end chip at the same price as the company's 8-core processor in 2017 is a great example of how AMD has continued to improve the quality ratio / price since the release of Ryzen. Obviously, we will have to wait for the processors to be verified, but this 12-core configuration is particularly important to exert serious pressure on the Core i9-9900K.
As for the 16-core rumor? Clearly is not there yet, but AMD should have the opportunity to publish one if it wishes. We expect the company to keep this option in reserve, potentially in response to any competitive initiative that Intel may be taking in the future to increase the number of its desktops. The clock speed of the engineering sample reported a few weeks ago is a bit low (3.3 GHz – 4.2 GHz), but engineering samples often have a clock a little lower than the final retail chips.
Once upon a time, motherboard chipset updates were major news – mainly because we were eagerly waiting for circuit engineers to fix what they could have broken into the previous hardware release. Fortunately, there are many less these days, but the flip side is that motherboards do not tend to evolve so quickly. A new PDF published by Biostar – probably before the company is supposed to be removed – suggests that AMD's next X570 chipset will resist this trend by offering support for several important features.
according to Biostar, its future Racing X570 GT8 will support PCIe 4.0, with two full-size slots configured electrically for x16 and x8 operation, and a third full-size slot with an x4 configuration suspended south of the southbridge. The three M.2 slots are described as being connected to the CPU and the southbridge, respectfully, using PCIe Gen 4 in all cases. The bandwidth is set to 32 Gbps, which corresponds to a PCIe 4.0 x4 link.
One thing to keep in mind, of course, is that SSD controllers will need to be updated to support PCIe Gen 4.0 themselves in order to take advantage of this bandwidth enhancement. A PCIe 3.0 M.2 drive plugged into a PCIe 4.0 port will still work perfectly, but will not take advantage of top speed. One of the theoretical benefits of switching to PCIe 4.0, however, is that motherboard manufacturers can offer the same high performance as a current X4 PCIe 3.0 solution while using half of the channels. (PCIe 3.0 Gen 3 offers 985 MB / s in each direction and per channel, while PCIe 4.0 increases it to 1,969 GB / s in each direction).
Update: Biostar's calculations do not work as shown. 32 Gbps is the bandwidth of the x4 link on PCIe 3.0, although 4.0 support is listed.
Offering both DDR4-4000 support and three M.2 slots on a card is a surprising twist. The AMD Ryzen 7 does not like high RAM clocks very much because of the need to use its Infinity fabric at half its speed. This means that you increase the DRAM clock, the IF clock also increases. According to some rumors, AMD would have set up a new divider that would run Infinity Fabric at a quarter of the indicated RAM clock (which corresponds to half of the current DRAM clock). In other words, instead of working at 2.5 GHz with DDR4-5000, the Infinity Fabric would return to 1.25 GHz – if this rumor is true. Alternatively, it is possible for Biostar to simply advertise high memory support without the corresponding improvements (as such) promised by AMD.
It would not be unprecedented for AMD to do this. We have not reached agreement with front-end bus multipliers for a long time, because motherboards no longer use FSBs, but some of them may remember the date your AGP clocks, PCI and ISA used multipliers derived from the FSB system. The use of multipliers was only intended to ensure that these low-speed buses remain within their own limits. One of the reasons why FSB overclocking was considered superior to overclocking by multiplier (but also more risky) is that its performance at faster clocks also improved transfer rates on ISA, PCI buses and AGP, as well as on the Northbridge memory controller clock and associated functions. RAM latencies.
But just because a motherboard, you can not stand that processors are practically reaching that frequency – and JEDEC has not yet released in DDR4 format above 3200 MHz. We would not consider this information as evidence of the high support of the third generation Ryzen RAM clock.
Separately, there is some unintentional humor on page 7 of the PDF.
AMD and Intel are definitely going both Be surprised by the fact that the X570 chipset supports Intel Core processors and, in a way, Intel B360. Interestingly, the X570 is the only new chipset on this list so far. According to rumors, AMD could have an update of the B550 chipset later this year, but for the moment, it seems that only the X570 is delivered at launch.
During AMD's deployment, you were asked about the amount of AMD products that have updated their product lines to strengthen their overall share of the x86 system market. The relationship between high product performance and actual market penetration is not necessarily linear and can be used depending on whether consumers actually purchase the products in question.
AMD's profits have improved since the start of Ryzen processors and the continuing shortage of Intel processors has opened up new opportunities for the company to sell in markets it has not targeted. AMD recently started selling Chromebooks, for example, by leveraging its old Bristol Ridge hardware to provide a platform considered equivalent to that sold by Intel in the same segment. But how was this done in terms of market share?
We now have data from Mercury Research via AMD, which means that the company has done quite well. The gains recorded over the past year have been substantial, while the quarterly trend of Q4 2018 in Q1 2019 is also positive.
In the first quarter of 2019, AMD gained one percentage point share of market share over the entire x86 market compared to the previous quarter and 4.7 percentage points of market share compared to the first quarter of 2018. This means that AMD has increased its market share by 1.54 times in just one year. substantial improvement for any business.
The gains were shared between desktops and laptops, but AMD was almost entirely a processor company for desktops and servers. The company gained 4.9 percentage points from year to year in desktop computers and slightly more than in laptops (5.1 percentage points). AMD still holds a larger share of office space than the laptop market, however.
The waiters are a bit of a puzzle, honestly. This is the second time that AMD has been transmitting Mercury Research data indicating one thing, then confidently asserting that its server share differs from that suggested by the Mercury Research chart. Earlier this year, AMD said it had taken about 5% of the market, while mercury was only 3.2%. This time, AMD reports that "the AMD server share (excluding IoT) was 2.9%, down 0.3 percentage points from the previous quarter and up 1.9 percentage points from the previous quarter. EPYC compared to Xeon DP server products remained unchanged at about 5% of device share. "
In other words, AMD estimates that it holds 2.9% of the total server market, but 5% of the dual-processor server market. Some of you may remember that AMD 's Opteron had targeted (and successfully captured) significant amounts of 4P server space, while we had not more than 1-2 processor cores per processor. Today, dual-socket server sales represent a much smaller market than single-socket servers, but AMD has not yet demonstrated the same strength in addressing this space in particular.
AMD does not seem to have reached its goal of penetrating the server market in single-digit averages, contrary to what Lisa Son had hoped for. However, the server space moves more slowly than mainstream PCs and at its own pace. AMD has demonstrated only one successful platform with Epyc, but server customers want to see what you can do in the longer term before committing to multi-year hardware marketing. All eyes will be on Rome, the 7th Epyc should debut later this year. If AMD's 7 nm hardware represents a significant advance over its predecessor, we can expect the server's market share to start growing as well. Bob Swan, CEO of Intel, said his company expects as much. During the first quarter 2019 teleconference, Swan acknowledged that I was expecting a more competitive environment on the server during the second half of the year after the Epyc update. AMD.
These figures may not be the most interesting data ever published by AMD, but they show the kind of positive movement that investors and customers want to see demonstrated by the company. AMD's best bet is to continue to consistently launch successful products, tackling markets in which Intel has long held more than 90% of the market. The markets for AI and machine learning are both key targets for AMD, but catching up with Intel and Nvidia – particularly Nvidia – is going to be a difficult lift. Big profits and regular execution are the solution to this problem, and AMD has done very well both lately. If Navi and Ryzen, a third generation group, launch smoothly this summer, the company should be well positioned to continue acquiring shares by the end of the year.