Platform Analysis

AMD provided the press with a block diagram of the Renoir APU at the time of the launch of the Zen 2 APUs. These APUs are monolithic dies, and true SoCs with all I/Os being sourced from the APU without a platform controller hub in the picture.

The above layout needs to be studied in conjunction with the design of the 4X4 BOX-4800U's motherboard. The various I/Os of the system (as well as internal components) are enabled using the following configuration.

An idea of the distribution of the various PCIe lanes can be obtained from the above diagram:

  • PCIe 3.0 x4 and a SATA port multiplexed behind the M.2 SSD slot (PCI Express x8 Bus #4)
  • PCIe 3.0 x1 for the Intel Wi-Fi 6 AX200 160 MHz WLAN card (PCI Express x1 Bus #3)
  • PCIe 3.0 x1 for the Realtek RTL8125 Gaming 2.5GbE Ethernet Controller (PCI Express x1 Bus #1)
  • PCIe 3.0 x1 for the Realtek RTL8168/RTL8111 GbE Ethernet Controller (PCI Express x1 Bus #2)

Other aspects of interest include the distribution of various USB ports - particularly in terms of bandwidth sharing. For example, we expect ports 'B' and 'C' behind the same root hub to share bandwidth. Note that the PCI Express x16 Bus #5 includes the two USB 3.1 controllers as well as the integrated GPU. The two Type-C ports in the front panel also act as conduits for two display outputs from the latter.

In the remainder of this review, we will first look at BAPCo's SYSmark 25, followed by various UL benchmarks and miscellaneous workloads. We also present some gaming benchmarks. A detailed look at the HTPC credentials of the system is followed by testing of the power consumption and thermal solution.

Introduction BAPCo SYSmark 25
Comments Locked

53 Comments

View All Comments

  • philehidiot - Wednesday, November 25, 2020 - link

    I've been drinking, it's 1900 and I've only eaten breakfast today so bear with me.

    I'm just astonished that we're now using 64GB DDR4 to eliminate any chance bottlenecking. My last system had 24GB DDR3 and I'm now on 16GB DDR4 with no memory concerns.

    As for *hic* SFF HTPC use, AMD has driver issues. I once said to a guy who deals in GPGPU stuff that "I don't have enough experience in drivers or coding to be able to say whether AMD's drivers are good or bad" and his response was "that's the same problem AMD have". If I'm using a PC for relaxing, having to debug or troubleshoot is not ideal. It's a shame, as noted in the conclusion here, that QA is often done on Intel systems and AMD don't get a look-in. Hopefully, as AMD become profitable and more performant they'll be able to convince people to QA on their hardware as well.

    They definitely need to sort out their drivers and assess the OEM support for other hardware. It's a sad fact that, since I dropped Intel for AMD (both CPU and GPU), I've had more crashes in a month than in years of Intel/Nvidia systems. I've lost more time to lost work than when I was using Windows 98 and CTRL+S after every change is back to being automatic.

    A few months ago I was trying to get 4K working on Netflix and their website specified only Nvidia card support or Intel iGPU decode support. Their support staff had never heard of the issue and so I went through the codecs supported by AMD and pointed out at least their hardware spec page should mention AMD once??? The guy couldn't help so he took it as a comment to escalate higher.

    /ramble. I need crisps.
  • StevoLincolnite - Wednesday, November 25, 2020 - link

    Got a secondary AMD rig that is 10+ years old and stays online for weeks at a time without a crash.

    Still runs games just fine, the more modern titles that are heavier threaded tend to run much better on the old Phenom 2 x6 and paired up with the Radeon RX 580 is a pretty capable 1080P rig.
    Will upgrade it when it's useless.
  • YB1064 - Wednesday, November 25, 2020 - link

    $900 is a bit too rich for my blood.

    BTW, what happened to the latest RTX 3xxx review? Did you guys give up on it?
  • Makaveli - Thursday, November 26, 2020 - link

    Apparently you didn't get the memo.
  • fcth - Friday, November 27, 2020 - link

    No explanation as to the delay (originally it was related to the fires in California, but it sounded like they were past that), but Ryan said a couple days ago that they are still getting caught up: https://twitter.com/RyanSmithAT/status/13317467171...
  • damianrobertjones - Wednesday, November 25, 2020 - link

    ...and Intel's driver support for Intel Nucs is hardly anything to scream about.
  • dontlistentome - Wednesday, November 25, 2020 - link

    Took Intel an age (over a year) to fix the BIOS updater on systems with Bitlocker - would fail to suspend the encryption so you'd be left with a machine that wouldn't boot, with a locked drive. Hope you kept the recovery key somewhere accessible.
  • Samus - Thursday, November 26, 2020 - link

    OMG I remember that bug, the worst part was there was no way to clone a drive to upgrade the SSD because you couldn't suspend bitlocker so you had to decrypt the drive which takes hours in some cases. 3-4 hour project just to upgrade an SSD...
  • Yorgos - Thursday, November 26, 2020 - link

    It took intel 3 or 4 years to fix the N2230 driver on linux.
  • bananaforscale - Thursday, November 26, 2020 - link

    I remember when Atoms first came and the NM10 chipset. It had crap support on Linux (didn't support resolutions beyond 800x600 for a while). Then there's Asus Transformer Book that has Intel hardware that's only supported under 32 bit Windows, so no upgrading to 64 bit (even if it would've been a bit silly in that particular case as it only had 2GB RAM). Linux is right out.

Log in

Don't have an account? Sign up now