Results 1 to 15 of 15

Thread: AMD's Ryzen 3000-supporting X570 chipset examined

  1. #1
    HEXUS.admin
    Join Date
    Apr 2005
    Posts
    30,229
    Thanks
    0
    Thanked
    1,940 times in 681 posts

    AMD's Ryzen 3000-supporting X570 chipset examined

    Lots of PCIe for ultra-fast storage.
    Read more.

  2. #2
    Registered+
    Join Date
    Sep 2009
    Posts
    68
    Thanks
    0
    Thanked
    2 times in 2 posts

    Re: AMD's Ryzen 3000-supporting X570 chipset examined

    In the first image "AMD X570: THE MOST MODERN I/O" see where that diagram says:
    x8 PCIe Gen4. Does that mean the X570 chipset will NOT support "4x4" AICs
    like the ASRock Ultra Quad M.2 AIC? The latter AIC needs a single x16 slot
    to "bifurcate" those lanes into 4 @ x4, hence the nomenclature "4x4".

  3. #3
    Registered+
    Join Date
    Sep 2009
    Posts
    68
    Thanks
    0
    Thanked
    2 times in 2 posts

    Re: AMD's Ryzen 3000-supporting X570 chipset examined

    P.S. See AORUS AIC GEN4 SSD here:
    https://hexus.net/tech/news/storage/131189-gigabyte-aorus-nvme-gen4-ssd-launched-copper-fin-heatsink/
    https://hexus.net/media/uploaded/2019/5/925eda35-423d-4905-8c83-ce2cceeec234.jpg

  4. #4
    Registered+
    Join Date
    Sep 2009
    Posts
    68
    Thanks
    0
    Thanked
    2 times in 2 posts

    Re: AMD's Ryzen 3000-supporting X570 chipset examined

    P.S. also here, with a few bullet points:
    https://pbs.twimg.com/media/D7oQOePUwAAPowF.jpg

  5. #5
    Senior Member
    Join Date
    Apr 2016
    Location
    UK
    Posts
    223
    Thanks
    64
    Thanked
    34 times in 22 posts
    • chinf's system
      • Motherboard:
      • 1: Asus M5A97 R2.0; 2: Asus M5A97 LE R2.0
      • CPU:
      • 1: AMD FX-8320E, Coolermaster Seidon V120 V2; 2: AMD PhenomII X3 710, Corsair H45
      • Memory:
      • 1: 16GB ECC DDR3L-1600; 2: 16GB ECC DDR3L-1600
      • Storage:
      • 1: Sandisk SSD, 4 WD, 1 Seagate, 1 HGST, 2 Toshiba HDDs; 2: Crucial SSD, 4 Samsung HDDs
      • Graphics card(s):
      • 1: HIS Radeon HD6850, Corsair H55 & NZXT G10; 2: MSI Radeon RX 480 8GB
      • PSU:
      • 1: be quiet! Straight Power 10 400W; 2: Antec EA500D
      • Case:
      • 1: Antec P182; 2: Antec Sonata
      • Operating System:
      • 1: Ubuntu 15.10 & 16.04; 2: Ubuntu 16.04 & Win10
      • Monitor(s):
      • Dell 24" @1920x1080, Iiyama 18" CRT

    Re: AMD's Ryzen 3000-supporting X570 chipset examined

    Quote Originally Posted by MRFS View Post
    In the first image "AMD X570: THE MOST MODERN I/O" see where that diagram says:
    x8 PCIe Gen4. Does that mean the X570 chipset will NOT support "4x4" AICs
    like the ASRock Ultra Quad M.2 AIC? The latter AIC needs a single x16 slot
    to "bifurcate" those lanes into 4 @ x4, hence the nomenclature "4x4".
    Good question. From a slot signalling perspective, you'd need to have a full 16-lane slot. Despite the red "x16 PCIe Gen 4" block in the second slide, I haven't yet seen any motherboards with such a configuration (full x16 slot via X570 chipset) - and even if there were any, it'd be pointless from a performance standpoint due to having to funnel all that bandwidth over a x4 Gen 4 link to the SoC.

    I think to max out M.2 provision on X570, one would first need to have an X570 board that offers 2x x16-physical slots wired to the CPU (not all do). These slots can be run as x8x8, and then a 2-way AIC to go into one of these alongside a GPU (assuming this is a desktop that needs a display card). Otherwise, there's the HEDT platforms.

  6. #6
    Registered+
    Join Date
    Sep 2009
    Posts
    68
    Thanks
    0
    Thanked
    2 times in 2 posts

    Re: AMD's Ryzen 3000-supporting X570 chipset examined

    Thanks for your feedback, chinf.

    On the upper left of that first graphic,
    see "PCIe 4.0 x4". That equals x16 lanes.

    Also, in the orange blocks further to the left,
    see "16x Graphics" and "4x Chipset Downlink"
    i.e. 20 total.

    So, the x16 "Graphics" slot might be available
    to support a bifurcated "4x4" AIC.

    It doesn't say how fast the "Chipset Downlink" is,
    but it would be reasonable for that Downlink
    to run at PCIe 4.0 speed e.g. see "1x4 NVMe".

  7. #7
    Registered+
    Join Date
    Apr 2017
    Posts
    32
    Thanks
    1
    Thanked
    0 times in 0 posts

    Re: AMD's Ryzen 3000-supporting X570 chipset examined

    I liked parts of it a lot, but ...

    "Here's a Corsair MP600 2TB NVMe M.2 SSD equipped with the latest Phison PS5016-E16 controller. It's benchmarked on an Asus Crosshair VIII Hero motherboard based on the X570 chipset. The left-hand benchmarks are run with the chipset's NVMe drive configuration set to PCIe 4 x4, whilst the right-hand benchmarks take in traditional PCIe 3 x4"u dont clearly say which nvme port is used. The more savvvy could guess, but not know.

    A comparison benchmark of the same drive on the chipset's nvme port would be enlightening.

  8. #8
    Registered+
    Join Date
    Sep 2009
    Posts
    68
    Thanks
    0
    Thanked
    2 times in 2 posts

    Re: AMD's Ryzen 3000-supporting X570 chipset examined

    > A comparison benchmark of the same drive on the chipset's nvme port would be enlightening.

    Indeed!

    The thrust of much of our activism over the past decade
    has been to stress the upstream potential of x16 slots,
    as compared to the limit necessarily imposed by a "DMI" bottleneck
    that supports only x4 PCIe lanes.

    That observation was quickly confirmed when prosumers
    tried to configure RAID arrays downstream of DMI links
    using very fast M.2 NVMe SSDs e.g. Samsung 970 Pro models.

    It was almost like the "100th Monkey" phenomenon:
    when one vendor finally released a "4x4" AIC with x16 edge connector,
    other vendors saw the light and jumped onboard.

    Let's hope the "auto detection" features of newer motherboards
    will become standard features for future builders who wish
    to do fresh OS installs to these 4x4 AICs.

    It's about time for the NVMe standard to embrace "native" support
    for all modern RAID modes, as did eventually happen with
    native SATA ports.

  9. #9
    Registered+
    Join Date
    Sep 2009
    Posts
    68
    Thanks
    0
    Thanked
    2 times in 2 posts

    Re: AMD's Ryzen 3000-supporting X570 chipset examined

    If anyone is interested, several months ago ASRock's Tech Support
    responded very quickly in answer to our question about their "4x4" support
    on the X399 and X399m motherboards. Here is the documentation
    they generously provided:
    http://supremelaw.org/systems/asrock/X399/

  10. Received thanks from:

    Millennium (11-11-2019)

  11. #10
    don't stock motherhoods
    Join Date
    Jun 2005
    Posts
    1,209
    Thanks
    748
    Thanked
    109 times in 94 posts
    • Millennium's system
      • Motherboard:
      • Asus Z170 Pro Gamer ATX
      • CPU:
      • Intel i5 6600K @ 4.5GHz 4 core
      • Memory:
      • Corsair VPX 3000 DDR4 (16, 4*4)
      • Storage:
      • 500gb 850 Evo sata3 SSD, 2*2TB Green 5900 Raid 0
      • Graphics card(s):
      • MSI 390 8gb
      • PSU:
      • toughpower 1kw
      • Case:
      • Zalman Z3 Plus
      • Operating System:
      • Windows 10 64bit
      • Monitor(s):
      • VIEWSONIC VG2401MH 144hz (Solid)
      • Internet:
      • Origin ADSL Broadband, not really recommended.

    Re: AMD's Ryzen 3000-supporting X570 chipset examined

    Quote Originally Posted by MRFS View Post
    If anyone is interested, several months ago ASRock's Tech Support
    responded very quickly in answer to our question about their "4x4" support
    on the X399 and X399m motherboards. Here is the documentation
    they generously provided:
    http://supremelaw.org/systems/asrock/X399/
    Stick around! I am curious as to your activism
    hexus trust : n(baby):n(lover):n(sky)|>P(Name)>>nopes

    how do you spend your time online? (Hexus link)

  12. #11
    Registered+
    Join Date
    Sep 2009
    Posts
    68
    Thanks
    0
    Thanked
    2 times in 2 posts

    Re: AMD's Ryzen 3000-supporting X570 chipset examined

    Gen3 / Gen4 / Gen5 PCI-Express permutations for x16 expansion slots


    Gen3: 4 x 2 x 8,000 / 8.125 = 7,877
    Gen3: 4 x 4 x 8,000 / 8.125 = 15,754
    Gen3: 8 x 2 x 8,000 / 8.125 = 15,754
    Gen3: 16 x 1 x 8,000 / 8.125 = 15,754

    Gen4: 4 x 2 x 16,000 / 8.125 = 15,754
    Gen4: 4 x 4 x 16,000 / 8.125 = 31,508
    Gen4: 8 x 2 x 16,000 / 8.125 = 31,508
    Gen4: 16 x 1 x 16,000 / 8.125 = 31,508

    Gen5: 4 x 2 x 32,000 / 8.125 = 31,508
    Gen5: 4 x 4 x 32,000 / 8.125 = 63,016
    Gen5: 8 x 2 x 32,000 / 8.125 = 63,016
    Gen5: 16 x 1 x 32,000 / 8.125 = 63,016


    There is a different frame encoding proposed
    for PCIe 5.0, but I haven't read any documentation
    of that future standard. Perhaps someone else here
    can share a few links to documents outlining
    that future encoding method.

  13. #12
    Registered+
    Join Date
    Sep 2009
    Posts
    68
    Thanks
    0
    Thanked
    2 times in 2 posts

    Re: AMD's Ryzen 3000-supporting X570 chipset examined

    I can think of at least two ways of encoding a frame
    in a way that is more efficient than the current 128b/130b "jumbo frame".
    An easy one is to drop the 8th bit from standard ASCII, which
    remains a 7-bit code (last time I looked). A more difficult one
    is to perform some form of compression on that 128b/130b "jumbo frame",
    such that each frame ends up being variable in length.
    I'm very curious to know the details about the proposed layout and
    logic of the Gen5 PCIe frame.

    The other "wild idea" I visualized recently is to map a 2-dimensional
    motherboard plane onto the six sides of a cube: each "side"
    would have its own set of PCIe expansion slots and DIMM slots.
    The literal "core" of this cube would be a future CPU connected
    to those slots via lanes that are designed to be as short as possible,
    to defeat the attenuation that occurs with very high data rates
    over very long transmission wires.

  14. #13
    root Member DanceswithUnix's Avatar
    Join Date
    Jan 2006
    Location
    In the middle of a core dump
    Posts
    11,033
    Thanks
    592
    Thanked
    1,191 times in 1,015 posts
    • DanceswithUnix's system
      • Motherboard:
      • Asus X470-PRO
      • CPU:
      • 3700X
      • Memory:
      • 16GB 3200MHz
      • Storage:
      • 1TB Linux, 1TB Games (Win 10)
      • Graphics card(s):
      • Asus Strix RX Vega 56
      • PSU:
      • 650W Corsair TX
      • Case:
      • Antec 300
      • Operating System:
      • Fedora 30 + Win 10 Pro 64 (yuk)
      • Monitor(s):
      • Benq XL2730Z 1440p + Samsung 2343BW 2048x1152
      • Internet:
      • Zen 80Mb/20Mb VDSL

    Re: AMD's Ryzen 3000-supporting X570 chipset examined

    Quote Originally Posted by MRFS View Post
    An easy one is to drop the 8th bit from standard ASCII, which
    remains a 7-bit code (last time I looked).
    I think you scuppered your own idea there, ASCII never had an 8th bit so you can't drop it Possibly that's why at the basic level the PCIe bus doesn't use ASCII, it needs 32 bit transfers as a minimum to comply with the underlying PCI spec, if 7 bits would do then I'm sure the people writing the spec would have used it.

  15. #14
    Registered+
    Join Date
    Sep 2009
    Posts
    68
    Thanks
    0
    Thanked
    2 times in 2 posts

    Re: AMD's Ryzen 3000-supporting X570 chipset examined

    We used to do systems programming on minicomputers:

    we had to write a quick "reformat" task that would reset the 8th bit
    if the ASCII source files had set the 8th bit. So, assuming the unused 8th bit
    is reset (equal to zero), here's one very rough "compression" example:

    01111111 01111111 01111111 (spaces shown only for clarity)
    when the unused 8th bit is dropped, becomes
    1111111 1111111 1111111

    Likewise, if the 8th bit is set, then:

    10000000 10000000 10000000
    becomes
    0000000 0000000 0000000

    Of course, such a protocol would need to be arbitrated
    successfully between the sender and receiver,
    before standard ASCII could be transmitted in that manner.

    In contrast, the binary strings above may NOT eliminate
    that 8th bit, because "binary" means that each binary digit
    is transmitted "as is".

    Hope this helps.

    /s/ Paul

  16. #15
    root Member DanceswithUnix's Avatar
    Join Date
    Jan 2006
    Location
    In the middle of a core dump
    Posts
    11,033
    Thanks
    592
    Thanked
    1,191 times in 1,015 posts
    • DanceswithUnix's system
      • Motherboard:
      • Asus X470-PRO
      • CPU:
      • 3700X
      • Memory:
      • 16GB 3200MHz
      • Storage:
      • 1TB Linux, 1TB Games (Win 10)
      • Graphics card(s):
      • Asus Strix RX Vega 56
      • PSU:
      • 650W Corsair TX
      • Case:
      • Antec 300
      • Operating System:
      • Fedora 30 + Win 10 Pro 64 (yuk)
      • Monitor(s):
      • Benq XL2730Z 1440p + Samsung 2343BW 2048x1152
      • Internet:
      • Zen 80Mb/20Mb VDSL

    Re: AMD's Ryzen 3000-supporting X570 chipset examined

    Quote Originally Posted by MRFS View Post
    We used to do systems programming on minicomputers:

    we had to write a quick "reformat" task that would reset the 8th bit
    Am interested in your top bit clearing, were you transferring data from a Prime computer by any chance? Those are the only ones I have come across that set the top bit rather than clear it when handling ascii bytes.

    Please believe me though, ASCII has no relevance at all to PCIe which almost entirely deals with binary graphics, network and filesystem data. Even modern characters are some mapping into the 32 bit Unicode space, usually using utf-8 which requires all 8 bits per byte.

Thread Information

Users Browsing this Thread

There are currently 1 users browsing this thread. (0 members and 1 guests)

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •