Setup Notes and Platform Analysis

The video below presents the entire gamut of available options in the BIOS of the 4X4 BOX-7735U/D5. Of particular interest is the 'CPU Operating Mode' under 'Advanced > CPU Configuration'. It is set to 'Normal' by default, corresponding to a TDP of 28W. Altering it to 'Performance' sets the fan speed to maximum irrespective of the actual load, but ekes out extra performance by pushing up the TDP to 42W.

The system is equipped with dual LAN ports backed up by Realtek controllers. Similar to previous 4X4 BOX systems, the 1GbE link comes with DASH support to make it easy for IT departments to deploy and manage the system with an out-of-band management interface. This support is disabled by default.

The block diagram below presents the overall high-speed I/O distribution.

The HDMI 2.1 port is actually driven from one of the DisplayPort outputs of the SoC with a Realtek RTD2175 protocol convertor on the board. The two Type-C ports in the front panel are both USB4 driven from the SoC in the same manner - a Realtek RTS5452E power delivery controller and a Kandou Technologies KB8002 retimer. One of the unfortunate aspects here is that this retimer doesn't support USB 3.2 Gen 2x2 (20 Gbps) operation, but all other interesting modes including full PCIe tunneling are supported. The rest of the platform components (including the network controllers, audio codec, etc.) are similar to the one in the previous 4X4 BOX series.

In today's review, we compare the 4X4 BOX-7735U/D5 and a host of other UCFF systems based on processors with TDPs ranging from 15W to 40W.

Comparative PC Configurations
Aspect ASRock 4X4 BOX-7735U (Performance)
CPU AMD Ryzen 7 7735U
Zen 3+ (Rembrandt R) 8C/16T, 2.7 - 4.75 GHz
TSMC 6nm, 16MB L3, 28W
Max / Target TDP : 50W / 42W
AMD Ryzen 7 7735U
Zen 3+ (Rembrandt R) 8C/16T, 2.7 - 4.75 GHz
TSMC 6nm, 16MB L3, 28W
Max / Target TDP : 50W / 42W
GPU AMD Radeon 680M (Rembrandt) - Integrated
(12 CUs @ 2.2 GHz)
AMD Radeon 680M (Rembrandt) - Integrated
(12 CUs @ 2.2 GHz)
RAM Kingston Fury KF548S38-8 DDR5-4800 SODIMM
38-38-38-70 @ 4800 MHz
2x8 GB
Kingston Fury KF548S38-8 DDR5-4800 SODIMM
38-38-38-70 @ 4800 MHz
2x8 GB
Storage Samsung PM9A1 MZVL2512HCJQ
(512 GB; M.2 2280 PCIe 4.0 x4 NVMe;)
(Samsung 6th Gen. V-NAND 128L (136T) 3D TLC; Samsung Elpis S4LV003 Controller; OEM version of 980 PRO)
Samsung PM9A1 MZVL2512HCJQ
(512 GB; M.2 2280 PCIe 4.0 x4 NVMe;)
(Samsung 6th Gen. V-NAND 128L (136T) 3D TLC; Samsung Elpis S4LV003 Controller; OEM version of 980 PRO)
Wi-Fi 1x 2.5 GbE RJ-45 (Realtek RTL8125)
1x GbE RJ-45 (Realtek RTL8111EPV)
Mediatek MT7922 (RZ616) Wi-Fi 6E (2x2 802.11ax - 1.9 Gbps)
1x 2.5 GbE RJ-45 (Realtek RTL8125)
1x GbE RJ-45 (Realtek RTL8111EPV)
Mediatek MT7922 (RZ616) Wi-Fi 6E (2x2 802.11ax - 1.9 Gbps)
Price (in USD, when built) (Street Pricing on April 17th, 2023)
US $630 (barebones)
USD 781 (as configured, no OS)
(Street Pricing on April 17th, 2023)
US $630 (barebones)
USD 781 (as configured, no OS)

The next few sections will deal with comparative benchmarks for the above systems.

Introduction and Product Impressions System Performance: UL and BAPCo Benchmarks
Comments Locked

35 Comments

View All Comments

  • 5080 - Thursday, April 6, 2023 - link

    Wondering if the 4X4 Box will be updated to the Ryzen 7 7840U or any of the other Phoenix based APU's in the near future.
  • TheinsanegamerN - Thursday, April 6, 2023 - link

    Yawn. Another ryzen 6000 rebrandeon product. Nobody cares about that, we want the zen 4+rDNA3 chips.

    And why oh why do these companies always put the thunderbolt on the front? Most thunderbolt devices are left plugged in, why would I want to dirty my desk with cables wrapped around the front of the PC?
  • lmcd - Thursday, April 6, 2023 - link

    It has all of the IO improvements relevant to this form factor. Zen 4 barely moved the needle compared to Zen 3, and RDNA3's biggest gains were in high-CU yields via chiplets. This form factor doesn't even benefit notably from the expected mild Ryzen 7x4x efficiency improvements.
  • TheinsanegamerN - Monday, April 10, 2023 - link

    20%+ increase in perf/watt, only ~5% loss going from 125 to 45 watts, and better clock scaling are "minor". Sure. Dont forget rDNA3, which is supposed to be a 40%+ improvement despite the same CU count.

    How is that rock you are living under?
  • lmcd - Wednesday, April 12, 2023 - link

    Zen 3 also scales down well, the difference is AMD wasn't advertising it.

    All of the improvements you listed are desktop 7000 platform vs desktop 5000 platform, and top SKU GPU vs top SKU GPU. Perf/watt improvements mostly came from the die shrink and improved bandwidth -- this platform has a die shrink and improved bandwidth, just the refined 6nm instead of newer 5nm. Clock scaling does not matter at these TDPs. RDNA3 performance literally cannot reach a 40% improvement over iso bandwidth (and there's no way the memory controller got completely reworked in a year's time).

    It feels like you eat up hype. Ryzen 7x4x will be nice, but its goal is to be the successor to Ryzen 5000, not Ryzen 6000. Ryzen 6000 is an up-to-date platform that has barely even stretched its legs.
  • evolucion8 - Tuesday, April 25, 2023 - link

    Intel in the other hand cant compete with AMD in the Sub 55W market with Zen 3+, let alone with Zen 4 ROFL.
  • meacupla - Friday, April 7, 2023 - link

    No, it's you who doesn't care.
    As it happens, this is a very fast APU, considering it is a rebranded 6000U series.
    Asrock somehow managed to tune the settings so that it is achieving extremely good results in gaming.

    As for front or rear ports, does it really matter? NUCs are best mounted on the backs of monitors anyways.
  • TheinsanegamerN - Monday, April 10, 2023 - link

    It achieves the same as any ryzen 6000 chip with DDR5. There's no magic tuning, asrock just stopped gimping.

    And yes, port placement does matter.
  • StevoLincolnite - Wednesday, April 12, 2023 - link

    No. Port placement doesn't matter on these devices. They are mounted behind monitors.
  • abufrejoval - Sunday, April 16, 2023 - link

    I own 5 NUCs, none was ever mounted to a monitor: they connect via a cascade of KVMs to a set of screens that have nothing mounted behind.

    I bought them for their low idle power and the small space they occupy under my desk. And they share that space with 6 Mini-ITX systems, four workstations and a set of semi-resident notebooks.

    While I value that at least their RAM and storage can be explanded, I'd still prefer Mini-ITX mainboards in 5L boxes, because generally they allow me to have or put anything inside (e.g. 10Gbit NICs) that I have to hang e.g. via Thunderbolt to the outside of a NUC.

    Unfortunately, getting "NUC power" in a Mini-ITX form factor has become nearly impossible, a very recent Erying G660 (https://wccftech.com/intel-alder-lake-h-laptop-cpu... which I've just added to my collection being one of the very few exceptions.

    BTW that board works rather well, at least after upgrading the Pico-PSU to 120 Watts, even if it only consumes 45 Watts on sustained peak loads.

    And thanks to a Noctua NH-L9i-17xx cooler it remains unnoticeable even under top load, something which the NUCs only ever achieve when you restrict their PL1/PL2/TAU settings to match their tiny fans.

    Just remember that personal computers have been loved for decades, exactly because people could use them in ways that their vendors never imagined.

Log in

Don't have an account? Sign up now