AMD's Ryzen CPU's (Ryzen/TR/Epyc) & Vega/Polaris GPU's

Discussion in 'Hardware Components and Aftermarket Upgrades' started by Rage Set, Dec 14, 2016.

  1. TANWare

    TANWare Just This Side of Senile, I think. Moderator

    Reputations:
    2,310
    Messages:
    8,796
    Likes Received:
    3,819
    Trophy Points:
    431
  2. hmscott

    hmscott Notebook Nobel Laureate

    Reputations:
    4,465
    Messages:
    15,665
    Likes Received:
    19,272
    Trophy Points:
    931
    The Fastest Linux Distribution For Ryzen: A 10-Way Linux OS Comparison On Ryzen 7 & Threadripper
    Written by Michael Larabel in Computers on 25 January 2018. Page 1 of 8. 36 Comments
    https://www.phoronix.com/scan.php?page=article&item=ryzen-linux-10way&num=1

    "While we frequently do Linux OS/distribution performance comparisons on the latest Intel desktop and server hardware, some requests came in recently about looking closer at the fastest Linux distribution(s) when running on AMD's Ryzen desktop processors. Here are benchmarks of ten popular Linux distributions tested out-of-the-box on Ryzen 7 1800X and Threadripper 1950X systems."
    "The Linux operating systems tested on these two current AMD platforms include Antergos, CentOS, Clear Linux, Debian, Fedora, Solus, Ubuntu LTS, Ubuntu, Void Linux, and openSUSE. Each operating system was cleanly installed and tested out-of-the-box for how Linux desktop users would find it when first deployed and applying all stable updates.

    Interested in seeing some fresh #Linux distribution comparison benchmarks about which Linux OS is the fastest on @AMDRyzen #Ryzen and #Threadripper systems? If so, which distros would you like to see compared?

    — Phoronix (@phoronix) January 20, 2018

    The choice of Linux distributions came from asking on Twitter about a Ryzen Linux distribution comparison. With updates, all of these distributions had at least minimal Retpoline support for Spectre v2 AMD mitigation but most (except Ubuntu) had the "full" support, a.k.a. being built with a patched compiler. KPTI (Kernel Page Table Isolation) wasn't needed and none of the tested distributions had inadvertently enabled it as Meltdown only affects Intel CPUs and all recent kernel patches treat it so accordingly. Some notes/summary on each of the tested distributions:

    Antergos 18.1 Rolling - This Arch-based Linux distribution is currently riding with the Linux 4.14.14-1 kernel, GNOME Shell 3.26.2, GCC 7.2.1, and an EXT4 file-system. One notable design choice of Antergos is using the CPUFreq Schedutil governor by default where as the other distributions were defaulting to ondemand or performance.

    CentOS 7 - CentOS 7 with updates is using the Linux 3.10.0-693.11.6 kernel, GNOME Shell 3.22.3, GCC 4.8.5, and an XFS file-system by default.

    [​IMG]
    Clear Linux 20500 - Intel's rolling-release distribution when tested had the Linux 4.14.14 kernel, GCC 7.2.1, and uses an EXT4 file-system by default. An additional item to note is that Clear Linux recently switched to using the KYBER I/O scheduler even on NVMe storage as used in today's testing where as the other Linux distributions all defaulted to "none" for their I/O scheduler choice while using the common EXT4 file-system.

    Debian Testing - Tracking the development of Debian 10 "Buster", the latest testing snapshot was used with the Linux 4.14.0-3 kernel, GNOME Shell 3.26.2, GCC 7.2.0 and an EXT4 file-system.
    [​IMG]
    Fedora Workstation 27 - This latest Fedora Workstation release has the Linux 4.14.14 kernel, EXT4, GNOME Shell 3.26 on Wayland, and GCC 7.2.1.
    [​IMG]
    Solus 3 - Solus with the Budgie desktop was included in this testing given it has been focusing on performance optimizations as well, some of which were pulled in from Clear Linux. Solus 3 with updates has the Linux 4.14.14 kernel, EXT4, and GCC 7.2.

    Ubuntu 16.04.3 LTS - The current Long-Term Support release of Ubuntu in its current HWE stack has Linux 4.13.0-31, GCC 5.4.0, and EXT4 atop a Unity 7.4 desktop.

    Ubuntu 17.10 - The current short-lived Ubuntu stable release with Linux 4.13.0-31, GNOME Shell 3.26.2 on Wayland, EXT4, and GCC 7.2.0.

    Void Linux - We benchmark Void Linux once in a while and saw some requests come in for Ryzen testing. This rolling-release Linux distribution with Xfce 4.12 had the Linux 4.14.15 kernel, GCC 7.2.0, and an EXT4 file-system.
    [​IMG]
    openSUSE Tumbleweed - SUSE's rolling-release is up to the Linux 4.14.14 kernel, KDE Plasma 5.11.95, GCC 7.2, and is using XFS for the home directory.

    For the duration of the benchmarking process, none of the distributions had any stability problems or other issues to note with the two Ryzen systems used for testing.
    [​IMG]
    Ryzen 7 1800X - The first system was the Ryzen 7 1800X with the MSI X370 XPOWER GAMING TITANIUM motherboard and BIOS 1.90. The system had 2 x 8GB DDR4-3200MHz memory, MSI Radeon RX 580 graphics card, and Corsair Force MP500 128GB NVMe SSD.
    [​IMG]
    Ryzen Threadripper 1950X - The Threadripper 1950X box had a Gigabyte X399 AORUS Gaming 7 motherboard with F3g BIOS. This system also had a Corsair Force MP500 128GB NVme SSD for storage, MSI Radeon R7 370 graphics card and 4 x 4GB DDR4-3600 memory.

    All ten of these Linux distribution benchmarks on the two AMD Ryzen systems were carried out in a fully-automated and reproducible manner using the open-source Phoronix Test Suite benchmarking software.
    [​IMG]
    First up was the basic SQLite embedded database benchmark for getting an idea of the rough I/O impact between distributions. Starting things off, Intel's Clear Linux distribution was the fastest on both Ryzen systems while CentOS 7 and Solus 3 were the slowest in this case.

    [​IMG]
    [​IMG]
    With CompileBench, Clear Linux was the fastest while CentOS 7 and Void Linux tended to be the slowest. Besides Clear Linux pulling in various performance optimizations and extra patches, as noted earlier, it did recently switch to using the Kyber I/O scheduler rather than no I/O scheduler for NVMe storage. Kyber is the recently mainlined I/O scheduler originally developed at Facebook.
    [​IMG]
    [​IMG]
    Moving onto more CPU focused tests, the Parboil OpenMP scientific test was the fastest on Clear Linux and Solus 3, which shares some of the performance optimizations of Clear Linux with Solus lead developer Ikey Doherty being a former Intel/Clear developer.
    [​IMG]
    In the MRI Gridding test the margins were closer while Debian Testing and openSUSE Tumbleweed this time were competing for the top spot.
    [​IMG]
    With the Rodinia OpenMP benchmark, Clear Linux was the fastest while CentOS 7 was notably slower with its stock compiler toolchain.
    [​IMG]
    And with HMMer, Clear Linux was again the fastest on these AMD Ryzen systems while Antergos 18.1 was the slowest.
    [​IMG]
    With the CPU-based TTSIOD phong renderer, Clear Linux had a narrow but measurable victory over the nine other distributions while Ubuntu 16.04.3 LTS this time was the slowest. Fedora, Solus, and Void Linux were other competitive performers in this benchmark.
    [​IMG]
    With x264 video encoding, Clear Linux once again came out in front... Yes, Intel's Linux distribution on AMD hardware. Ubuntu 16.04.3 and 17.10 as well as Antergos 18.1-Rolling were notably slower than the others.
    [​IMG]
    The ebizzy benchmark aims to resemble web-server workloads. On the Ryzen 7 1800X box, Antergos 18.1 had a narrow path to victory. The results on the Threadripper 1950X system were quite close among most of the distributions tested, but Clear Linux attained another win. Ubuntu 16.04.3 LTS was noticeably slower on both systems than the others tested.
    [​IMG]
    When it came to measuring the time needed to compile a Linux x86_64 default configuration kernel, CentOS, Clear Linux, and Ubuntu 16.04.3 LTS were all nearly tied for first while Void Linux was the slowest.
    [​IMG]
    C-Ray ray-tracing was the fastest on Clear Linux, Ubuntu 16.04.3 LTS, and CentOS 7 while the other seven Linux distributions were running slower but at around the same speed.
    [​IMG]
    The FLAC audio encoding results were mostly flat but if counting outright wins/losses, Clear Linux again had the fastest result while CentOS 7 was the slowest.
    [​IMG]
    The FFmpeg results were also close but with Clear, Fedora, and Solus competing for the fastest while Antergos 18.1 was noticeably slower.
    [​IMG]
    [​IMG]
    With the Hackbench kernel benchmark, Void Linux was substantially slower than all other tested kernel configurations while Antergos and CentOS won the cake this time.
    [​IMG]
    For the total boot time, Clear Linux booted the fastest on the Ryzen 7 1800X system while on Threadripper 1950X, Solus 3 was the fastest followed by Antergos and then Clear Linux. On both systems, CentOS 7 and openSUSE Tumbleweed were the slowest. Void Linux numbers were not available as it's one of the few distributions not using systemd.
    [​IMG]
    The PyBench results for looking at the Python performance was the fastest on Clear Linux followed immediately by Debian Testing and Ubuntu 17.10. CentOS 7, Solus 3, and Void Linux were among the slowest.
    [​IMG]
    Lastly was PHPBench with Antergos 18.1 and Clear Linux performing for the top spot while CentOS 7 with its PHP5 stack by default being the slowest.

    Of all the benchmarks ran, Clear Linux 20500 was the front-runner. On the Ryzen 7 1800X system, Clear Linux won 65% of the time followed by Antergos 18.1 with 15% of the wins and Solus in third. On this Ryzen 7 1800X box, CentOS 7 was the slowest 38% of the time given its older but vetted Enterprise Linux 7 stack.

    With the Threadripper 1950X system, Clear Linux was the fastest 57% of the time followed by Solus at 15% and Antergos 18.1 at 7%. CentOS 7 here was the slowest with it coming in last 46% of the time.

    So as shown previously in past benchmarks, while Clear Linux is a project out of Intel's Open-Source Technology Center, even when running on modern AMD x86_64 hardware it still packs a mighty performance punch thanks to its aggressive compiler defaults, FDO / LTO / FMV / other compiler performance techniques, various backported patches for performance, and many other smaller optimizations. Hopefully more Linux distributions will take note in 2018 and work on similar performance aspirations, in part to make up for some of the performance losses incurred by the Meltdown and Spectre mitigation techniques. As shown when testing on Intel hardware with KPTI and full Meltdown protection, the fully-patched Clear Linux system can generally still outperform distributions prior to their Spectre/Meltdown penalties. Solus 3 and Antergos also deserve shout-outs for their strong performance on these two AMD Ryzen systems.

    For those wondering about the OpenGL graphics performance, there are results on the following page. It wasn't a main focus for this testing since it's not really Ryzen-specific and mostly comes down to a race of who has the most-updated Mesa/kernel. But there are a few OpenGL results for those interested.
    [​IMG]
    Lastly onto the extra results featuring a few OpenGL tests of the Ryzen 7 1800X with the Radeon RX 580 on supported platforms. As just stated, the OpenGL tests weren't a main focus due to not being Ryzen-specific, mostly just a straight-forward Mesa/kernel race, and this testing already being quite lengthy... Due to all the time involved, just some standalone Linux OpenGL games/benchmarks were used rather than setting up a Steam environment on each platform, etc.
    [​IMG]
    [​IMG]
    Antergos 18.1 with its up-to-date stack tended to be in front.
    [​IMG]
    [​IMG]
    Even with the demanding Unigine tests, not too much variation, hence the lack of focus on graphics/gaming in this comparison.
    [​IMG]
    CentOS 7 tends to be the slowest due to its dated graphics drivers.
    [​IMG]
    If you enjoyed this comparison, consider showing your support by joining Phoronix Premium to support future large Linux hardware/software comparisons.
     
    Last edited: Feb 26, 2018
    ajc9988 likes this.
  3. hmscott

    hmscott Notebook Nobel Laureate

    Reputations:
    4,465
    Messages:
    15,665
    Likes Received:
    19,272
    Trophy Points:
    931
    AMD A320 or B350 For Raven Ridge? Buying The Right Motherboard
     
    Raiderman likes this.
  4. hmscott

    hmscott Notebook Nobel Laureate

    Reputations:
    4,465
    Messages:
    15,665
    Likes Received:
    19,272
    Trophy Points:
    931
    Ryzen (1000-Series) Overclocking Guide
    Want to overclock your Ryzen Threadripper CPU? Here's an OC guide.
    https://www.tweaktown.com/guides/8506/ryzen-1000-series-overclocking-guide/index.html

    "AMD's Ryzen CPUs have been out for almost a year, and it's now time for us to put out an easy to understand and easy to follow overclocking guide for the CPU. The AM4 platform has matured a lot, and so things are much more stable and overclocking is easier now than ever.

    AMD offers two ways to overclock their Ryzen CPUs; through the motherboard's UEFI and through AMD's Ryzen Master application. We are going to focus on the UEFI as AMD has done an excellent job of covering how to overclock with their software. If you aren't okay with overclocking in the UEFI you can read AMD's Ryzen Master OC Guide."
    8506_02_ryzen-1000-series-overclocking-guide_full.png
    "The flow chart above is pretty basic but spells out some things you should know before you begin. At around 95C the CPU will throttle, so you want to stay at least 10-15C below that for safety. CPU core voltage maximum is 1.45v, and SoC voltage is 1.2v.

    AMD's Ryzen CPUs do not run off a VID table; they run between 0.2-1.5v depending on load and environmental characteristics, but when you overclock, you should not set anything above 1.5v (I doubt you could even cool a CPU at that voltage). You should expect a 3.9-4.1GHz overclock.

    While temperature is typically your limiter, in this case, you will not be going over 4.1GHz on ambient cooling solutions. Some Thread Ripper CPUs can hit 4.2GHz, but Thread Ripper is made out of the top 5% of all Ryzen dies. I always aim for 4GHz, but I have seen some of my CPUs not hit 4.0GHz.

    The cool thing about the new processors is that their multipliers can be increased in 0.25 increments. AMD's AGESA code is the basic code AMD gives its partners to build a UEFI. AMD has made big leaps with code updates, especially on the memory overclocking side, and they publish a lot of information on their blogs as you can see here.

    Disclaimer Overclocking your CPU technically can damage your CPU. TweakTown and the writer of this guide take zero responsibility if you damage or kill your CPU. There is also a chance that AMD will not replace a CPU damaged by overclocking. AMD states that "The limited warranty does not cover damages due to external causes, including improper use, problems with electrical power, accident, neglect, alteration, repair, improper installation, or improper testing".

    Have you Overclocked Before?
    If you have overclocked before and understand hardware selection and the basics of overclocking, you should skip to the next page. The first part of this guide is for those who want to know what to do before overclocking.

    Where do I start?
    Ryzen Overclocking Guide Systems

    CPU: All Ryzen SKUs can overclock

    Motherboard: You will need to use an X370 or B350 motherboard to overclock your CPU. A-series motherboards are designed to maintain standard reliability, so overclocking is locked on those motherboards. When looking for a motherboard to overclock with, the main limitation on the motherboard will be the VRM. Our reviews extensively cover motherboard VRM and we thermal test every motherboard, so look for our reviews.

    DRAM: AMD and motherboard vendors have greatly improved DDR4 compatibility and speed potential, but we still recommend buying a kit off your motherboard's Qualified Vendors List (QVL). AMD's AGESA 1.0.0.6 code greatly increased DRAM overclocking potential, and increased maximum memory multiplier up to 40x, but so far most people are seeing 3200-3600Mhz as their limitation.

    There are also many vendors that have memory certified for AMD Ryzen; I highly recommend those kits. Also, you can look at the memory we use in our Ryzen reviews, as that exact kit has been validated by AMD to work at the speed of the profile of the kit. Dual-rank modules are harder to find these days, but if you do come into possession of them don't expect overclocks similar to single rank kits. Single rank kits overclock much easier. Dual rank kits are typically double-sided modules.

    Cooler: High-end air coolers or all-in-one watercoolers are recommended. The maximum frequency of your CPU will probably be 4.1GHz, and the way voltage scales on the platform means that increasing cooling to crazy levels probably won't get you up to 4.1GHz. That means that even mid-range air coolers are good enough for most Ryzen overclocking.

    PSU: I would leave about 150-200W aside for a nice overclock on an Ryzen 7 1800X.

    You enter the BIOS/UEFI by hitting delete or F2 (on most boards) during boot up. For most boards you have basic and advanced modes, I always skip to the advanced mode and tend to navigate with the keyboard. To enter a setting you either type (or delete and then type), use +/- keys, or you click and scroll. Then you have to "Save & Exit" the BIOS/UEFI for the settings to apply.

    CPU Multiplier
    Every brand with an AM4 motherboard is included in all of our images of settings to alter. We have done this to include all overclockers.
    [​IMG]
    Some motherboards require you to choose a mode on how to overclock. ASRock might also include CPU Voltage in multiple areas; you should only need to change it in one place. ASUS allows you to choose to overclock with D.O.C.P/XMP mode, which is how we do it on their motherboards. It automatically takes up your RAM to its profile settings and then allows you to set CPU ratio. On Biostar motherboards at the time we tested, you could not easily input a multiplier. Instead, you had to rely on core FID and DID values, and Core VID could be used to change the VCore.

    On the Biostar motherboard you can increase FID and leave DID at 8, and you will see the greyed out frequency increase or decrease based on what you do. On the GIGABYTE board you see above we just change CPU clock ratio. MSI has a normal and expert mode, we just use expert mode, but you can also use normal. Remember you can change the multiplier in 0.25x increments so that you can increase CPU frequency in 25Mhz steps. Default voltage when you overclock should be 1.3625v, and you can change it, but on many motherboards I just use the 1.35v that comes with default VID. Coincidentally my nice 1800X likes 1.35v for 4GHz.
    [​IMG]
    There two ways you might be able to increase your overclock if you reach a thermal or another limit. You can disable Simultaneous Multi-Threading (SMT) which will remove the two threads per core feature of the CPU. So if you disable SMT on a 1800x (8 cores 16 threads), you will get only eight cores with eight threads. You can also turn off cores, and that can be done through Downcore Control. The good news here is that all motherboard vendors have decided to call these settings the same thing.
    [​IMG]
    Some motherboards allow users to access AMD's CBS menu. Inside the menu, the settings are standardized as this is part of AMD's core part of the UEFI. If you navigate to AMD CBSZen Common OptionsCustom Core Pstates and you accept the warning, you can actually change individual core P states and FID, DID, and VID. You can see the resulting frequency and voltage. I believe that VID is in hexadecimal. I wouldn't overclock through this menu, but there has been a lot of discussions online about it.
    [​IMG]
    The main voltage you need to change to overclock the CPU is the CPU Core voltage; most vendors call this VCore. Don't set this over 1.45v for 24/7 use, although cooling the CPU at that high of a voltage is pretty though. I like to stay below 1.4v. By default the CPU VCore should go to 1.3625 when you overclock, you might want to change the voltage manually, so you aren't over or undervolting the CPU. You also can change the CPU SoC voltage, and that should increase memory overclocking potential. Default SoC voltage is 0.99v, and AMD recommends no more than 1.2v. There are also other voltages you might need to change, for instance on most motherboards you need to increase DRAM voltage to what your sticks want.
    [​IMG]
    Depending on your motherboard you might also have VRM/PWM settings that control the external VRM's PWM controller. These settings can be used to stabilize a fluctuating voltage and increase power limits. Load Line Calibration (LLC) is used to stabilize fluctuating voltage, switching frequency determines the aggressiveness of the VRM (higher is more aggressive but less efficient), and over current and voltage protects can be increased to maximize output. You typically don't need to touch any of these settings other than LLC. CPU SoC LLC is called VAXG on GIGABYTE boards and NB on MSI motherboards.
    [​IMG]
    Some motherboards might have a more basic VRM, and on those motherboards, you might not have the ability to set the whole voltage and instead just an offset. The offset is an amount added to the base voltage of the CPU, so that should be between 1.35 and 1.3625 when you change the CPU multiplier. The SoC should be 0.99v by default so your offset can be added to that.

    Power Saving Settings
    [​IMG]
    Most motherboard vendors have decided to call AMD's advanced CPU features the same thing. So AMD's Cool'n'Quiet can be disabled if you want to disable the CPU's low-power c-states and you can disable C6 as well if you want. On Intel platforms, many users disable virtualization to help increase stability (I tend not to bother), and on AMD SVM is virtualization.
    [​IMG]
    Vendors have taken it upon themselves to make up their own memory profile names. XMP (ASRock/GIGABYTE), D.O.C.P., and A-XMP are all the same thing. They enable a profile built in the memory module to automatically overclock it. It's a good idea on all motherboards to manually set DRAM voltage, and sometimes you need to set the DRAM voltage for each channel. DRAM voltage, frequency, and primary timings can be found on the DRAM module's sticker.
    [​IMG]
    While AGESA 1.0.0.6 brought on a ton of more timings, many users might not understand the complexity of those timings and might just opt to change primary timings you see above. On some motherboards memory profiles might not work, in that case, you need to manually set your memory frequency and primary timings.
    [​IMG]
    If you want to tweak the new timings and other features like gear-down mode, you can read AMD's block post found here. There is also a DDR4 PHY voltage, I see it more often on Thread Ripper motherboards, it's called CLDO_VDDP, and AMD states this voltage can solve some memory frequency holes. It's an offset voltage you can take up and down (both can be useful). It should be configured to DIMM voltage -0.1v, but you aren't to exceed 1.05v offset on this.

    Stability Testing

    [​IMG]
    I prefer running lighter (HandBrake) and more obsolete (IntelBurnTest, IBT) stress tests for quick testing to see how stable the CPU is while making one tweak after another. However, just running these types of tests and passing doesn't mean your CPU is 24/7 stable. Memtest is one of the most popular tests for testing memory stability. From my experience, if you can pass Handbrake encoding with a large 4K video or IBT, you aren't far off from your voltage required for 24/7 stability, but just running something like HandBrake or IBT isn't enough to say your CPU is totally stable. I like Handbrake for quick testing because it spits out a performance figure in FPS in its log and you can see throttling with that number, and it is a real-world test and uses AVX. However, AVX is not a huge issue for Ryzen CPUs because they don't have large dedicated AVX units built into them. I like Intel Burn Test because it does get your CPU very hot and if your system is very unstable it will freeze, but if it is mid-stable, it will error out while HandBrake will just skip to the end of the queue.
    [​IMG]
    Some users will just go to the strong stress testers like Prime95, and run it for a day or two. I mainly use three stability testing programs; HandBrake encoding of a 4K video (very quick but shows performance and uses AVX), Intel Burn Test with a decent chunk of memory usage (has AVX and can be run however long you like), and Prime95 blend one day and small FFTs for another (latest version with AVX). Prime95 has a few different tests you can run. The default test is a blend test, which tests most everything, but isn't going to demolish CPU core instabilities as Small FFTs would. AIDA64 also has a built-in test, and you can choose what you test (core, FPU, etc.) but I would increase the amount of RAM used in that test as the default value is a bit low. AIDA64 is considered a "safe" test by many, as it's not designed to beat your CPU up, but many believe you can't become a man until you get a really good beating. Over time, it's possible for certain stress testing programs to do better than others, and the best place to find this information is in the stress testing threads on overclocking forums.
    [​IMG]
    When you stress test, it's also a great idea to keep a monitoring program up and running. I like HWinfo, as it monitors a ton of metrics and is often updated. I wouldn't keep up multiple monitoring programs. That means I highly recommend using a program like HWinfo and disabling or uninstalling any other software like motherboard monitoring software. It's also a good idea to not have a CPUz tab up either, as it polls the CPU a lot and HWinfo already reports on CPU speed. You might also want to set your Windows Power Plan to high-performance. Overall, in the end, some people recommend up to 2 days of stress testing to ensure your system is totally stable. The only thing to keep in mind in the long term is that any overclocking and hardcore stress testing can negatively impact the lifespan of your components, but overclocking is fun and can increase performance with ease."
     
    Last edited: Feb 27, 2018
    TANWare likes this.
  5. TANWare

    TANWare Just This Side of Senile, I think. Moderator

    Reputations:
    2,310
    Messages:
    8,796
    Likes Received:
    3,819
    Trophy Points:
    431
    I'd say that is about right for what to expect from at least the TR series. I would have liked to have seen more but TBH at 16 cores 4.0 GHZ is a lot of computing power.
     
    Raiderman, ajc9988 and hmscott like this.
  6. TANWare

    TANWare Just This Side of Senile, I think. Moderator

    Reputations:
    2,310
    Messages:
    8,796
    Likes Received:
    3,819
    Trophy Points:
    431
  7. hmscott

    hmscott Notebook Nobel Laureate

    Reputations:
    4,465
    Messages:
    15,665
    Likes Received:
    19,272
    Trophy Points:
    931
    Peter Chambers On The 2018 AMD Ryzen Mobile Laptops

    The 2018 Acer Ryzen Mobile Laptops Revealed!

    The 2018 AMD Ryzen Mobile Laptops First Look!
     
    Raiderman likes this.
  8. TANWare

    TANWare Just This Side of Senile, I think. Moderator

    Reputations:
    2,310
    Messages:
    8,796
    Likes Received:
    3,819
    Trophy Points:
    431
    https://overclock3d.net/news/cpu_ma...potted_-_how_much_faster_is_it_than_a_1700x/1
    https://www.techporn.ph/amd-celebrates-ryzens-one-year-anniversary/
    http://thenokiablog.com/2018/03/02/...ews-and-features-everything-you-need-to-know/

    Get out your salt shakers for this next one. Spells doom and glom for Intel and nothing but sunshine for AMD. My primary complaints are claiming a 40% performance hit from Meltdown and 12nm possibly eliminating Intel IPC advantage. With the later it may help some but I highly doubt eliminate it.

    https://www.bitcoinisle.com/2018/03/01/amds-growing-cpu-advantage-over-intel/
     
    Last edited: Mar 3, 2018
    Raiderman and jaybee83 like this.
  9. hmscott

    hmscott Notebook Nobel Laureate

    Reputations:
    4,465
    Messages:
    15,665
    Likes Received:
    19,272
    Trophy Points:
    931
    Sapphire Nitro+ Radeon RX VEGA 64 -- how good is it?
     
    Raiderman likes this.
  10. Deks

    Deks Notebook Virtuoso

    Reputations:
    983
    Messages:
    3,971
    Likes Received:
    1,167
    Trophy Points:
    231
    Here's the extremetech article on the subject:
    https://www.extremetech.com/computi...n-significant-clock-boost-upcoming-ryzen-cpus

    So, we see a supposed 300 MhZ update over 1700x on the baseline.
    If those numbers are accurate, the 2800x in comparison might be clocked at 3.9/4.0 GhZ and boost to about 4.4-4.6 GhZ - and if the manuf. process is indeed more suited for higher clocks as the technical info from Glofo says... then people will probably have bigger overclocking headroom allowing them to reach Intel's frequencies.

    Though, I suspect that if we want AMD to match Intel in overclocked frequencies (which isn't exactly a fair comparison)... I suspect that 2600x will be the needed CPU to accomplish that due to its smaller core count (which should allow it to clock higher).

    The IPC differential between Intel and AMD is only 5%. Its possible AMD might have addressed this issue in addition to optimizing latencies with Infinity Fabric with the refresh... though, historically, we didn't get IPC increases with refreshes... only clock speed increases.

    As I said before, we'll need to wait and see.
     
    Raiderman likes this.
Loading...

Share This Page