Quote:While "Summit Ridge" is the combination of two "Zen" CCX (quad-core CPU complex) units making up an 8-core CPU die that lacks integrated graphics, the "Raven Ridge" silicon combines one "Zen" CCX with an integrated graphics core based on the "Vega" architecture. AMD's new Infinity Fabric interconnect ferries data between the CCX and the iGPU, and not an internal PCIe link. The CCX houses four "Zen" CPU cores with 64 KB of L1I cache, 32 KB of L1D cache, 512 KB of dedicated L2 cache, and 4 MB of L3 cache shared between the four cores.

The integrated graphics core is a different beast. It features similar (albeit scaled-down) front-end and back-ends from the "Vega 10" silicon, a similar video engine, and an SIMD area with 10 "Vega" next-gen compute units (NGCUs). This works out to a stream processor count of 640. Other key specifications include 40 TMUs, and 16 ROPs.

The video engine is now extremely capable, supporting hardware-accelerated decoding of CODECs such as VP9 10-bpc and HEVC 10-bpc at frame-rates of up to 240 for 1080p, and 60 for 4K UHD. It can also encode H.265 8-bpc at frame-rates of up to 120 at 1080p, and 30 at 4K UHD. You finally get to use the display connectors on your socket AM4 motherboards, as the iGPU supports DisplayPort 1.4 and HDMI 2.0b, with resolutions of up to 3840 x 2160 @ 60 Hz with HDR, 1440p @ 144 Hz, and 1080p @ 240 Hz.
...
AMD seems so have increased the amount of power-gating on its silicon. Disabled or idling components triggered by lower power-states, are now power-gated (their power-supply cut off), and not clock-gated (their clock cadence cut-off). The chip is peppered with multiple LDO (low-dropout regulator) regions for the CCX, iGPU, and uncore regions, with a common VDD package rail for both the off-chip (on motherboard) and on-chip voltage controllers.

Quote:One chip that isn't on the road map that MoePC got ahold of is an as-yet-unknown AMD "Fenghuang" APU with "15FF" graphics. This product found its way into the SiSoft results database recently. The 15FF IGP apparently has 28 Radeon compute units for a total of 1792 shader processors, accompanied by 2 GB of an unknown type of VRAM. Other data that made its way into the database would appear to be spurious, like the 555-MHz clock speed, 16kB of L2 cache, and a supposed 32-bit path to memory. There isn't much in the way of details about this chip, but it definitely appears to be a prototype. Still, if it is real, Fenghuang's IGP would represent a considerable increase in graphics resources over today's Vega 8 and Vega 10 IGPs. The power needed to support that much graphics horsepower could peg this chip as a desktop part, but we won't know either way until AMD offers more details, if it ever does.

Quote:The entire Ryzen "Raven Ridge" APU lineup was put through 3DMark 11 "Performance" preset, by someone with access to all of them. The 2400G leads the pack with 5,162 points, and a graphics score of 5,042 points. The 2200G, which features 512 stream processors, and lacks SMT, manages 4,151 points, with 3,950 points graphics score. The 2400G scores somewhere between the desktop RX 550 and the RX 560, which makes it possible for you to run "Player Unknown's Battlegrounds" at 900p or even 1080p with some details dialed down.

(02-12-2018, 10:34 PM)SteelCrysis Wrote: http://www.tomshardware.com/reviews/amd-ryzen-5-2400g-zen-vega-cpu-gpu,5467-11.html
https://www.extremetech.com/computing/263683-amd-ryzen-5-2400g-review-best-blend-cpu-gpu-performance-weve-ever-seen
https://www.techpowerup.com/reviews/AMD/Ryzen_5_2400G_Vega_11/11.html
https://www.techpowerup.com/reviews/AMD/Ryzen_5_2400G_Vega_11/12.html
https://www.techpowerup.com/241444/amd-ryzen-raven-ridge-comes-with-a-limited-pcie-interface
The summary of all of these is that Ryzen APUs will play 720p to 1080p at low to high settings, and still won't come close to an RX 460 or a GTX 1050, though they are roughly equal to a GT 1030 or an RX 550. On top of all of this, Ryzen APUs only have 8x PCIe 3.0 lanes available for graphics cards.

Quote:It is also interesting that AMD uses heat-conducting paste instead of solder between its Raven Ridge dies and heat spreaders. However, with an average power dissipation of less than 100W, this cost-cutting measure is probably tolerable for everyday operation. Of course, we also ran a series of more demanding workloads to tax both processors. It comes as little surprise that we figured out how to get the 2400G to throttle. We weren't expecting, however, to get its Radeon Vega Graphics engine stuck that way.
...
All of this wouldn’t really be worth writing about if it wasn’t for the fact that Ryzen keeps on throttling, even after the conditions that caused throttling in the first place are relaxed. A reboot is necessary to reset the chip's operating parameters. There’s no rhyme or reason to this, and AMD can't explain it.

The good news, however, is that both of AMD's new Ryzen processors can be cooled without any issues using the company's bundled Wraith Stealth cooler. We used a number of especially demanding workloads to push these chips as hard as possible. But that's above and beyond what they'll encounter outside of the lab.

Quote:Our thermal data yields a couple of surprises. First, the paste that AMD uses between its die and heat spreader performs better than we believe a lot of enthusiasts were expecting. Second, it'd be hard for anyone to improve on the stock configuration, since the paste is applied so thinly to the heat spreader during production.

Gluing the heat spreader under pressure furthermore ensures an optimal result. When we carefully removed the original layer of thermal paste, we took note of how much AMD used and replicated its effort with our own high-quality stuff. Upon removing the heat spreader a second time, we confirmed that our application looked just about as perfect as AMD's.

Nevertheless, we did not succeed in achieving a noticeable advantage over the original paste's performance. Not even a burn-in over several hours helped to significantly improve our findings. The hotter Ryzen 5 2400G became, the smaller the already tiny difference got. In addition, when we used the boxed cooler, both applications of thermal paste failed to keep AMD's chip from hitting its thermal limit with Prime95 and MSI Kombustor running. Naturally, then, replacing the stock paste with another silicone-based paste is pointless. AMD makes the best of its position by achieving optimal performance within the framework of industrial mass production and pressure to cut costs.

(02-19-2018, 10:35 PM)SickBeast Wrote: Ryzen runs hot with the IGP, just as I would expect with Vega integrated.  No overclocking improvement, either.  For that I guess we have to wait for Zen+, if it even happens.

(02-17-2018, 04:43 AM)SteelCrysis Wrote: AMD goes even farther to fix UEFI issues, promising to send affected users new Ryzen APUs for free that aren't affected by the UEFI issues: https://www.extremetech.com/computing/264097-amd-shipping-free-processors-customers-address-apu-firmware-update-issue

(02-19-2018, 11:05 PM)SteelCrysis Wrote:

(02-19-2018, 10:35 PM)SickBeast Wrote: Ryzen runs hot with the IGP, just as I would expect with Vega integrated.  No overclocking improvement, either.  For that I guess we have to wait for Zen+, if it even happens.

What's worse is that APUs, and gaming on integrated graphics, are stuck at the same place where they were when AMD introduced the APU: insufficient for 1080p on modern games.

(02-20-2018, 02:05 AM)SickBeast Wrote: AMD must have signed an agreement with Sony stating that they were not allowed to release the PS4 APU for PC use.  That would have made an amazing entry level PC gaming product.

(02-20-2018, 02:17 AM)SteelCrysis Wrote:

(02-20-2018, 02:05 AM)SickBeast Wrote: AMD must have signed an agreement with Sony stating that they were not allowed to release the PS4 APU for PC use.  That would have made an amazing entry level PC gaming product.

Yes, but Intel had no such restrictions. Broadwell was an amazing first step for iGPU development. If Intel had stuck with it, they could have started to compete with AMD and Nvidia without having to actually develop their own graphics cards. But Intel threw away all of that progress, and now they're having to hire Koduri just to get somewhere.

(02-20-2018, 03:11 AM)SickBeast Wrote: Can they not build upon what they already have with Broadwell?

Quote:We don’t necessarily recommend using a Ryzen 5 2400G for Windows Mixed Reality. Because even though headset prices have fallen (in some cases below the $250 mark), and performance graphics card pricing remains sky high, most users looking to get into the VR space probably don't want to invest several hundred dollars on a system with serious limitations. But our experience with AMD's Raven Ridge chip opened our eyes to the possibility that we may soon have access to chips with integrated graphics that are powerful enough for the demanding workload of immersive VR experiences. We wouldn’t be surprised if AMD gives its second-generation Ryzen APUs enough oomph to drive high-quality VR content. At that point, PC-based VR gaming will have a better chance of gaining mainstream traction than in the current world where highly overpriced graphics cards are required.

This experiment also gives us hope that we’ll eventually see VR on the Xbox One platform. Two years ago, Microsoft introduced the Xbox Scorpio project and said that the new console would support virtual reality. When the company dropped the new console (now called Xbox One X), it stopped talking about VR on the Xbox platform. But we always expected that the Windows Mixed Reality platform and the Xbox One platforms would eventually come together to offer a mainstream living room VR gaming experience. Microsoft's Windows Mixed Reality runs surprisingly well on a mainstream chip like the Ryzen 5 2400G without official support or optimization. So we have to image that the two companies, working together, could deliver a more-satisfying experience with the powerful SoC in the Xbox One X.

Quote:AMD's latest Radeon Software Adrenalin Edition 18.6.1 Beta, which is available now, lacks Raven Ridge APU support. Driver support for the APUs are limited to WHQL releases only, as noted by an AMD representative on the Overclockers UK forum. Currently AMD is set to use a three month release cycle for APU drivers. Understandably, this has caused some concern with the latest driver to offer support for the Raven Ridge APUs being the Adrenalin Edition 18.5.1 driver released in May. The only good news here is the limited driver releases allow AMD to further optimize their costs in regards to testing and qualification.

Limited or outdated drivers, with such a long period between releases, means games could perform sub-optimally on AMD's latest and greatest APUs. Worse yet, consumers could be stuck waiting three months for an updated driver. Even then, if a problem arises and is a fringe issue, fixes could take even longer. Essentially Raven Ridge owners are being left out in the cold to some extent in regards to hot-fixes and performance improvements. This makes AMD's Raven Ridge APUs with built in VEGA graphics for both desktops and mobile systems a bit less appealing. This issue is further exacerbated by the fact Intel's Kaby Lake G series which also features AMD's VEGA graphics has seen a new driver released that is based on the 18.6.1 driver.

Quote:The company revealed the new Raven Ridge processors while it was outlining its progress towards its 25X20 goal, which states the the company will offer 25X more power efficiency by 2020. AMD touts the new "Raven Ridge 2018" processors as more efficient than the existing models, which again furthers the theory that these processors could be 7nm process. We are digging for more information and will update this article as necessary.

Quote:AMD released two patches for its open-source AMDGPU DRM Linux driver yesterday. The latest patches add support for the red chipmaker's unreleased Picasso and Raven Ridge 2018 APUs (accelerated processing units).
...
A purported AMD roadmap Uruguayan media outlet Informática Cero published pointed to Picasso as the code name for Raven Ridge's successor. Another leaked PowerPoint slide seemed to confirm that Picasso employs the same architecture as Raven Ridge, which might be true as AMD confirmed in one of yesterday's patches that Picasso is a new APU similar to Raven. If that is the case, we wouldn't expect any groundbreaking performance over the current generation of APUs. However, Informática Cero's unconfirmed information does point to performance and power improvements, which could possibly mean a die shrink. Our assumption is that Picasso will either be manufactured under Global Foundry's 12nm node or TSMC's 7nm node.

Picasso processors will be available in desktop and mobile variants. In the case of desktops, the APUs should slot perfectly into the AM4 socket, which AMD plans to retain until 2020. If there are no setbacks, Picasso should arrive in 2019.

Quote:AMD listed the Ryzen 7 2800H and the Ryzen 5 2600H to its website. These new processors bring the inherent goodness of the Raven Ridge architecture, found in the Ryzen 5 2400G and the Ryzen 3 2200G, to gaming notebooks. As such, these processors come with AMD's Zen compute cores paired with the Vega graphics architecture, and they are also AMD's first processors to support DDR4-3200 as a base specification.
...
The move to support for DDR4-3200 is an encouraging sign, as we've shown how readily the Raven Ridge parts respond to improved memory performance. Pair that with Vega's higher clock speeds and these should be pretty powerful chips for the gaming laptop segment.

Quote:So just how is it that the enabling of a tiny few components or increasing nominal clock speeds have such a tremendous impact on TDP? Perhaps there are other under-the-hood settings these chips have that make them more eager than their U-series siblings. Also, nominal clocks are clock speeds that each of the four CPU cores on the chip are guaranteed to run at, beyond which, depending on a number of factors, the PrecisionBoost algorithm awards higher clocks. On the U-series chips, PrecisionBoost is extremely conservative with boost clocks. Even so, could clock speeds really have such a profound impact on TDP? Join the debate in the comments below.

Quote:Anandtech has received confirmation from AMD that it will not bring an APU to market based on the Matisse design in 2019. This doesn’t mean AMD has no plans to bring a 7nm APU to market — only that the CPU in question will be a custom design. Because APUs are intended to serve lower-priced markets, AMD makes different design decisions concerning them and tends to position them for different markets.
...
In other news, AMD has stated it expects the TDP range of Matisse products to be the same as the current lineup of Ryzen CPUs. This implies CPU TDPs ranging from 35W to 105W. It is not clear if that applies to the theoretical bit of empty space or not. Maintaining full socket compatibility with AM4 is important for AMD’s long-term compatibility pledge, but squeezing a high-performance 16-core CPU into a 105W TDP while maintaining high clock speeds might be a bit tricky. Then again, there’s clearly room for something. We’ll have to wait and see.

Quote:Asus and MSI likely have similar reasons behind their decisions: BIOS chip storage capacity. As AMD's CPU collection on the AM4 platform grows, so does the size of BIOSes which must accommodate all AM4 CPUs. Lower-end boards typically have only 8MB of storage for BIOSes, whereas higher end boards get 16MB. Supporting Bristol Ridge (the A-Series APUs), Summit Ridge (Ryzen 1000), Pinnacle Ridge (Ryzen 2000), and Raven Ridge (Ryzen 2000 APUs) all at the same time has become difficult due to chip capacity constraints. Now AMD is adding the Matisse processors to the mix.

To deal with this, vendors have two options. Asus apparently chose one, which is to simply stop supporting updates for cheaper A320 motherboards that didn't have the storage necessary to store bigger BIOSes. MSI has chosen the other, to remove support for Bristol Ridge or other previous architectures to make room for new additions. This could make more sense than what Asus has done since the A-Series isn't exactly a popular choice for AM4 motherboards, and it's unlikely many people will be angry with MSI's decision to stop supporting these last-gen APUs.

Either way, though, these decisions are impacting AMD's promise to support the AM4 platform until 2020. It's nice to have CPUs that work on the same socket every generation, but if some motherboards can only support some of those generations at a time, then it's not entirely different than what Intel did with its 300 series of motherboards: keeping the socket but not allowing previous-gen CPUs to work on the new chipset, or new CPUs to work on the old chipsets.

Quote:We found the chip running on a Colorful CVN X570 V20 motherboard, meaning this was our first encounter with a Ryzen 3000 series chip running on an X570 motherboard. Here we can see the chip notching a single-core score of 162 and a multi-core score of 712 in CineBench R15, but bear in mind the chip could be a pre-production version.

However, although both the Ryzen 5 3400G and Ryzen 3 3200G come to market with 3000-series branding, they shouldn't be confused with the Ryzen 3000-series desktop processors. The APUs, codenamed Picasso, feature the Zen+ microarchitecture and 12nm node while the desktop chips, codenamed Matisse, are based on Zen 2 and the 7nm manufacturing process.
...
The four-core eight-thread chip runs at a 3.8 / 4.2 GHz base/boost, a nice increase over the Ryzen 5 2400G. The sample was running with DDR4-2400 memory settings, but we're sure that isn't the official memory speed. The reduced memory speed throws off the benchmarks we ran, meaning that higher performance could be possible with higher memory transfer rates. Unfortunately, we were unable to adjust the settings in the BIOS. From previous listings, it appears the chip supports DDR4-3200 memory.

Quote:AMD has announced the latest Ryzen 3 3200G and Ryzen 5 3400G APUs (Accelerated Processing Units) at the "Next Horizon Gaming" event. As you can probably guess by the Ryzen 3000-series branding, the new APUs are the generational successors to the Ryzen 3 2200G and Ryzen 5 2400G parts.
...
The Ryzen 3 3200G employs Radeon Vega 8 graphics clocked at 1,250 MHz, which is 150 MHz faster than the Ryzen 3 2200G's iGPU (Integrated Graphics Processing Unit). AMD includes the Wraith Stealth CPU cooler with the Ryzen 3 3200G. Despite all the aformentioned improvements, the Ryzen 3 3200G still has a 65W TDP (thermal design power) and comes with a friendly $99 price tag.
...
While it retains the same Radeon RX Vega 11 graphics, the Ryzen 5 3400G's iGPU operates with a 1,400 MHz base clock. Once again, we're looking at a 150 MHz improvement over the existing Ryzen 5 2400G. AMD's Wraith Spire CPU cooler is included with this SKU. Surprisingly, the Ryzen 5 3400G is expected to cost $149, which is even cheaper than the Ryzen 5 2400G that debuted at $170.

According to AMD's internal benchmarks, its Radeon RX Vega 11 is significantly faster than Intel's UHD Graphics 630 iGPU. From a gaming perspective, the Ryzen 5 3400G is capable of providing smooth gameplay above 30 frames per second at 1920x1080 resolution. However, AMD didn't specify which graphic presets were used in the tests.

Quote:Motherboard manufacturer ASRock has listed the specifications for eight unnanounced AMD desktop APUs (Accelerated Processing Units) in the company's processor support list database. The next-generation APUs, codenamed Picasso, are set to replace the current Raven Ridge family.

Quote:Renoir is the supposed successor to AMD's current Picasso APUs. Picasso is built with the 12nm manufacturing process and sports Zen+ cores and Vega graphics. Renoir is rumored to make the transition to the 7nm process node and feature the latest Zen 2 microarchitecture. The general expectation was that Renoir would make use of the new Navi graphics solution. However, recent updates to AMD's open-source AMDGPU display driver have seemingly dispelled this rumor.
...
An unconfirmed AMD roadmap, which was leaked by Uruguayan media Informática Cero last year, points to Renoir landing in 2020. If the launch year is accurate, Renoir will compete against Intel's 10nm Ice Lake chips that feature the Gen11 graphics solution. In Intel's own benchmarks, its quad-core Ice Lake-U (ICL-U) chip barely manages to pull out a win against the similar quad-core Ryzen 7 3700U Picasso APU. It's important to note that the Ice Lake-U processor was running with LPDDR4X-3733 memory while the Ryzen part was paired with DDR4-2400 memory. In any case, we expect Renoir to close the small gap, if not completely blow Ice Lake out of the water.

Quote:Recent Linux driver patches deployed on August 28 seemingly hint that next-generation AMD APUs, codenamed Renoir, will come with support for LPDDR4X-4266 memory.

Memory plays a crucial role with APUs. Tests have shown that faster memory can often improve the chips' performance, especially in a gaming environment. For reference, AMD's current mobile Picasso U-and H-series APUs officially support DDR4-2400 memory, which is a bit underwhelming considering that Intel's Ice Lake laptop processors support both DDR4-3200 and LPDDR4-3733 formats. If the latest Linux patches are to be trusted, Renoir could show up with an improved IMC (integrated memory controller) that can accommodate LPDDR4X-4266 memory modules.
...
The word on the street is that Renoir is scheduled to arrive in 2020 to replace Picasso; however, AMD hasn't confirmed this.

Quote:Someone has posted data claiming to be 3DMark 11 benchmark results for the upcoming AMD Renoir APUs on Reddit. According to the Redditor, these are results that aren't accessible by the general public because they're set to "private."

There are three results which seem to show Renoir APUs running on a 'Celadon-RN' motherboard. We suspect that Celadon could be the name of the platform for Renoir. Unfortunately, the results don't reveal APU model names or core counts. We have no way of telling if the results are supposed to represent one, two or three different APUs. All things considered, we should take these mysterious results with a grain of salt.
...
Based on the results, the tested chips seem to have low base clocks, but this is normal considering this would be very early silicon. If these really were Renoir chips tested, AMD probably still has a lot of tweaking to do. Alternatively, the low clock speeds could be a sign that Renoir will come with higher core counts. The current generation of AMD APUs, Picasso, tops out at four cores.

Another thing that stands out with these alleged benchmarks is the iGPU clock speed. One of the setups had the iGPU at 1.5 GHz. Picasso APUs' iGPU maxes out at 1.4 GHz. Renoir potentially surpassing Picasso at this early stage makes AMD's next-generation APU look promising.

Quote:After initial rumors said that the AMD’s upcoming line of Zen 2-based APUs , codenamed Renoir, would arrive with integrated Vega 10 graphics, the latest rumor points to the chips featuring Vega 12 and potentially Vega 13 and Vega 15 integrated GPUs.

The rumor comes from well-known hardware leaker Komachi_Ensaka on Twitter , who reportedly observed multiple Renoir listings with B12 included in their names, which could point which to 12 Compute Unit (CU) GPUs.

Hardware enthusiast Locuza also noted that due to AMD’s use of the 7nm process for the next-generation APUs, it may also be possible to integrate Vega 13 or even Vega 15 GPUs in there, as the newer processors allow for denser designs.