AMD Ryzen 6000 RDNA 2 iGPU Smashes Iris Xe DG1, GeForce MX350 In New Benchmarks

AMD Ryzen Processor
AMD Ryzen Processor (Image credit: AMD)

It looks like Intel isn't the only one with a processor that supports DDR5 memory. A new UserBenchmark submission (courtesy of Benchleaks) reveals that AMD isn't far behind from embracing the new wave of DDR5 memory kits.

AMD's Ryzen 6000 (reportedly codenamed Rembrandt) is expected to carry Zen 3+ cores that are complemented by Navi 2 (RDNA 2) graphics. From a processing perspective, the chips should retain the same core count as the Ryzen 5000 (Cezanne) series, but feature improved clock speeds. Rembrandt plays a very important role in AMD's future lineup as the APUs are rumored to be the first parts to leverage Navi 2 graphics as well as support for DDR5 memory.

The user ran the benchmark on Corsair's Xenomorph device, which could be an upcoming mini-PC from the manufacturer unless the company plans to venture into the gaming laptop market. UserBenchmark also specified the FP7 socket that should be the new platform for the Ryzen 6000 series. The Ryzen 6000 processor from the UserBenchmark entry lacks a name so we'll have to refer to it by its 100-000000518-41_N OPN code for now. But if we have to speculate, it could be the Ryzen 7 6800H or Ryzen 9 6900HS/HX. UserBenchmark detected the unreleased AMD chip with an eight-core and 16-thread configuration after all. 

According to the report, the Zen 3+ APU sports a 3.1 GHz base clock and 3.9 GHz boost clock, but the OPN code points to 4.1 GHz. The processor is probably an engineering sample so the final specifications can vary. Remember to take the benchmark results with a truckload of salt since UserBenchmark is known for being a biased benchmark tool. Additionally, the Ryzen 6000 is a single submission, while the scores from the other processors in the comparison are averaged results from different users.

AMD Ryzen 6000 CPU Benchmarks

Swipe to scroll horizontally
Header Cell - Column 0 Core i7-11800HRyzen 7 5800H100-000000518-41_N
Memory Score70 Pts76.6 Pts67.9 Pts
1-Core Score161 Pts132 Pts111 Pts
2-Core Score315 Pts257 Pts228 Pts
4-Core Score592 Pts496 Pts399 Pts
8-Core Score995 Pts875 Pts740 Pts

The Ryzen 7 5800H (Cezanne) outperformed the mysterious Ryzen 6000 processor by up to 19% in the single-core test and up to 18% in the octa-core test. The Ryzen 6000's low scores can be explained by the possibility of the chip being an early engineering sample or the fact that it was paired with single stick of DDR5-4800 C40 SO-DIMM memory module so dual-channel operation wasn't enabled. The latter also contributed to the higher memory latency on the Ryzen 6000 processor. The Ryzen 7 5800H configuration showed 13% lower memory latency in comparison to the Ryzen 6000 chip.

Apparently, the Core i7-11800H (Tiger Lake) is superior to the Ryzen 6000 sample. The Intel chip's single-and octa-core scores were 45% and 34% higher than the Zen 3+ APU, respectively. However, the Core i7-11800H did have 3% higher memory latency.

AMD Ryzen 6000 iGPU Benchmarks

Swipe to scroll horizontally
Header Cell - Column 0 1CFA 0004GeForce MX350Iris Xe DG1
Lighting29.5 FPS24.8 FPS23.6 FPS
Reflection131 FPS26.6 FPS27.2 FPS
MRender9.7 FPS17.3 FPS29.2 FPS
Gravity33.7 FPS24.2 FPS31.6 FPS

On the graphics end, the Ryzen 6000 APU appeared with a RDNA 2 iGPU (1CFA 0004) with 512MB of shared memory. It's a shame that the software couldn't pick up the number of compute units or their clock speeds.

The RDNA 2 iGPU surpassed Intel's Iris Xe DG1 in the lighting, reflection and gravity  tests by 25%, 382% and 7%, respectively. However, the Iris Xe DG1 beat the RDNA 2 iGPU in the MRender benchmark by a 201% margin.

It would seem that the MRender is the only workload that gave the RDNA 2 iGPU a hard time. Nvidia's GeForce MX350 put up a 78% higher MRender score than the RDNA 2 iGPU. However, the latter destroyed the GeForce MX350 by 19% in the lighting test, 392% in the reflection test and 39% in the gravity test.

According to a previously leaked roadmap, Ryzen 6000 is allegedly scheduled for a 2022 debut. If the rumors around the hardware circles are true, Ryzen 6000 should already be in mass production with a potential launch for the first half of next year.

Zhiye Liu
RAM Reviewer and News Editor

Zhiye Liu is a Freelance News Writer at Tom’s Hardware US. Although he loves everything that’s hardware, he has a soft spot for CPUs, GPUs, and RAM.

  • deesider
    Encouraging, but rather weak compared to Apple's new M1X. The eventual Mac-mini version would make a fantastic mini game machine - just a pity there won't be any games to run on it.
    Reply
  • Makaveli
    Need to fix this statement.

    "UserBenchmark isn't known for being a biased benchmark tool. "
    That should either saying is known

    or being a non biased benchmark tool.
    Reply
  • srimasis
    And I was hoping to get Gtx1050ti levels of performance.
    Reply
  • srimasis
    deesider said:
    Encouraging, but rather weak compared to Apple's new M1X. The eventual Mac-mini version would make a fantastic mini game machine - just a pity there won't be any games to run on it.
    Unfortunately AMD will never make a chip faster than M1 max because they don't need to. They will keep making the same old 8 core chips for gamers. And people who need a USD 3000+ laptop for productivity would rather go for a desktop with a threadripper than a laptop. Unless they are tied to the MAC os.
    Reply
  • spongiemaster
    So the Intel iGPU and RDNA 2 perform pretty similarly in 2 tests. Intel wins one by 201% and AMD wins the last by 382%? That doesn't look like a very useful benchmark suite. You're not going to see real world gaming benchmark suites showing such wild swings.
    Reply
  • TerryLaze
    Makaveli said:
    Need to fix this statement.

    "UserBenchmark isn't known for being a biased benchmark tool. "
    That should either saying is known

    or being a non biased benchmark tool.
    Yes because CPU and GPU is the same thing.
    Also they still give you the real CPU numbers right at the beginning so you can judge by yourself if you have any knowledge on the subject, they are just telling you their opinion on what to expect if you are a normal user that isn't heavily into productivity.
    Reply
  • ChaosFenix
    The fact that this silicon performed as well as it did in any of these benchmarks is promising for the silicon. Both the MX350 and DG1 are dedicated graphics(albeit low-power) solutions being compared to an iGPU. The iGPU is being heavily throttled here by a single stick of RAM as well since memory bandwidth is shared with the CPU on integrated graphics. As for other caveats you have that the CPU clock may be throttled here and driver optimizations haven't been done. The only thing that may lower performance going forward is really down to power as we don't know if the test system was running at 10W or 65W. If it was on the low end then depending on they system you may see more performance but it this was run on the high end you may see this lower in an actual laptop.

    Edit: As the author said we also don't know how many CUs were enabled for the run but if any of them were disabled we also may see more performance in the future.
    Reply
  • hotaru251
    srimasis said:
    They will keep making the same old 8 core chips for gamers.
    until market has use for mroe cores....there IS no reason to push more.

    8cores is basically all u need for non workstation/server (and they have those options in TR and epyc)


    and lookign at the difference between zen1, zen, and zen3? those are nice improvements. (lot mroe than we've seen from intel's new gen chips up til 12th gen)


    srimasis said:
    AMD will never make a chip faster than M1 max
    in single core usage*

    Ryzen still wrecks any apple silicon in multithread performance....oh and doesnt carry apple tax nor suffer from not being user serviced.
    Reply
  • Rdslw
    deesider said:
    Encouraging, but rather weak compared to Apple's new M1X. The eventual Mac-mini version would make a fantastic mini game machine - just a pity there won't be any games to run on it.
    The problem with ARM is game engines depend on x86 and if you would migrate them to arm FPS would be horrible. Games made for arm will be mostly mobile and shared arch, which means they wont be hard to run.
    but I guess you could run fortnite mobile with 200fps here. Oh wait, you wont because apple hate epic games.


    hotaru251 said:
    Ryzen still wrecks any apple silicon in multithread performance....oh and doesnt carry apple tax nor suffer from not being user serviced.
    Yes in desktop, no in mobile workstation. With Video works on laptop, windows is the weakpoint of ryzen. If you run on battery, windows throttles so much, you won't get much done there, while apple would happily work.
    zen 3d will even out to m1 cores in single threaded, and as zen4 seems to be doubling multithreaded performance, if it includes that 3d cache as well, then amd will leave everyone else in the dust.

    Problem for apple is that arm chip is that Apple just used every single option they had. Even if its winning now, they have no growth options anymore, over incremental 20% each stepping each 2 years
    Reply
  • renz496
    Rdslw said:
    The problem with ARM is game engines depend on x86 and if you would migrate them to arm FPS would be horrible.

    this is a misconception. the FPS would be slow because of RAW performance on ARM CPU not because it was ARM vs x86. if the ARM core are being designed with high performance along side high power requirement in mind it should have no issue matching x86 performance.
    Reply