1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

PCIe Gen 4 SSDs perfom Better on Intel PCIe Gen 3.0 vs AMD Gen4/3

Discussion in 'Processors and motherboards AMD' started by MegaFalloutFan, Jul 22, 2019.

  1. MegaFalloutFan

    MegaFalloutFan Master Guru

    Messages:
    615
    Likes Received:
    72
    GPU:
    RTX 2080Ti 11Gb
    So this comes as huge surprise, GEN 4.0 NVMe ssds been benchmarks to perform better on Intel PCIe Gen 3 vs AMD Gen 4.0 and AMD Gen 3.0.
    Basically AMD is on the BOTTOM of performance across both Gen 4.0 and Gen 3.0.

    Only in Synthetic Reads MP600 performed better on x570

    For example in Crystal Disk Mark 4K Q32T1, Intel Gen 3.0 beat x570 Gen 4.0 by Significant margin of 40MB/s

    The benchmarks are on Polish web site but its easy to understand, just start from Page one and click next to see all of them

    https://www.purepc.pl/pamieci_masow..._test_ssd_na_platformach_intel_i_amd?page=0,3
    [​IMG]

    https://********/FKY4XzX/1.jpghttps://********/ryqHkHg/2.jpg
    [​IMG] [​IMG] [​IMG]



    P.S. As one that intent to buy 16 core CPU and top mobo like MSI Prestige or ASUS Hero, i been dissapointed so far with Ryzen 3000, it has SERIOUS issues
    1) [reported by Derbauer and basically every owner of 3900x and 3800x] with single core boost, even with all the available updated, these CPUs wont boost to Advertised 4.5 and 4.6Ghz speed in Single thread tests [only spiking here and there].
    2) Many people reported of being stuck at base clocks.
    3)This PCIe Gen 4 speed issue, its outrageous that Intel with its PCIe Gen3 beats AMD with its high end PCIe Gen 4 using AMD Optimized SSDs [the first gen PCIe 4 SSD Controller was created with AMDs help, they sais so during Ryzen announcement]


     
    Last edited: Jul 22, 2019
  2. Astyanax

    Astyanax Ancient Guru

    Messages:
    3,037
    Likes Received:
    783
    GPU:
    GTX 1080ti
    Theres nothing surprising about these results especially since they are using agesa 1.0.0.2.

    can't really get full performance when you're resubmitting data via transactional retries to get around the bus parity errors.
     
    Last edited: Jul 23, 2019
    Jackalito likes this.
  3. MegaFalloutFan

    MegaFalloutFan Master Guru

    Messages:
    615
    Likes Received:
    72
    GPU:
    RTX 2080Ti 11Gb
    Are you talking about the video I Posted? Its unrelated, its another issue posted by different people.
     
  4. RaV[666]

    RaV[666] New Member

    Messages:
    8
    Likes Received:
    3
    GPU:
    MSI HD6950 TwinFrozr 2GB

  5. MegaFalloutFan

    MegaFalloutFan Master Guru

    Messages:
    615
    Likes Received:
    72
    GPU:
    RTX 2080Ti 11Gb
    The moment you blame someone for shilling to another team, you invalidate yourself and make yourself look like a die hard fanboy.
    I posted a link to all tests, they have 19 pages of tests and as I MENTIONED in my post
    CPU has nothing to do with PCIe speeds, Intel had numerous security fixes that lowered its NVMe speeds across the board, yet it beats both Ryzen platforms [x570 and last gen] and in PCIe Gen 4 specifically.
    And in Synthetic Benchmarks 3600 is faster then 8700

    The only available GEN4 NVMe controller was developed with direct AMD involvement.

    Lets not forget all the otehr issues Ryzen x570 platform has: Cant boost to advertised speeds on higher end CPUs, stuck with high voltage settings in windows, on older chipsests some CPUs stuck at base clocks.
    They need to fix all these bugs, before they turn into permanent "features"

    P.S. BTW, I dug up TweakTown Ryzen gen 1 NVMe benchmarks and looks like Ryzen had issue with PCIe speed since day one.
     
  6. Astyanax

    Astyanax Ancient Guru

    Messages:
    3,037
    Likes Received:
    783
    GPU:
    GTX 1080ti
    This is false, the cpu definitely plays a part in random read performance.
     
  7. MegaFalloutFan

    MegaFalloutFan Master Guru

    Messages:
    615
    Likes Received:
    72
    GPU:
    RTX 2080Ti 11Gb
    This is true, the difference is less then 2%, its irrelevant.
     
  8. toasty

    toasty Member Guru

    Messages:
    108
    Likes Received:
    3
    GPU:
    1080Ti 2GHz <1v
    The nvme thing is not something I personally care about. I chose not to go for the overpriced - actively cooled X570 boards and instead picked up a decent X470. My Gen3 M.2 hits the same sort of speeds in here as it did with a 7700K/Z270-F and is more than fast enough for typical usage, I'll look at Gen4 when we have GPU's fast enough to care about it.

    The rated boost clock is however something I do personally care about and feel slightly cheated by AMD on this.
    They went balls to the wall on voltage to achieve the clocks they print on the box and while mine can hit it's rated speed, only a couple of cores are even actually capable to begin with and then it's only for <1s duration; it's useless. It's like their ultra sophisticated boost algorithm goes "if running at <95°C and low power/load then request 1.5v and boost to the rated boost clock" then immediately proceeds to crap itself and throttle. The issue is exacerbated by the fact that windows and the programs we run have no idea which are the "golden cores" to take advantage of.

    The actual boost is really more like 150MHz-200MHz less (when not restricted by cooling and/or power delivery, more if so).

    The stupid thing about this? I can achieve higher usable all core clock speeds than stock can do in even single thread, and lower temps and get better performance by tuning it myself. You can overdo it and run into clock stretching if you try to clock too high with too little voltage, it might not seem unstable or throttle; but performance will suffer.
     
  9. MegaFalloutFan

    MegaFalloutFan Master Guru

    Messages:
    615
    Likes Received:
    72
    GPU:
    RTX 2080Ti 11Gb
    What you describe is NOT getting the boost, its simple Spiking, Boosting is like 9900K does it goes to 5Ghz until it hits Thermals and then clocks down and the thermal limit can be easily disabled to make it permanently boost.
    So spiking to 4.6Ghz for one second, is not really how it supposed to be.
    I care about PCIe, because i have 3 PCIE storage devices, i want to do 2TB RAID0 NVMe, i have Intel 480GB Optane for OS drive, i want the 16 core CPU and ASUS hero or MSI Prestige, but of course i need all of it to work properly no PCIe bugs or boost issues.
    X570 has another bug, WHEA errors if you have GEN 3 PCIe devices installed and file corruption [I posted another thread here with images]


    P.S. I also assumed that new Ryzen chips will be able to do like Intel does, to lock them into all core permanent boost, but 4.6Ghz on 3900x all core is dream.
    My only hope is that from what i heard the 3950x should be the best CPU out there, the best silicone so they might boost better, maybe AMD takes all the AAA core and puts aside for the 16C CPU? I hope they are
     
  10. RaV[666]

    RaV[666] New Member

    Messages:
    8
    Likes Received:
    3
    GPU:
    MSI HD6950 TwinFrozr 2GB
    Well, no. Their tests are cpu heavy and not only 8700K is just a much faster cpu at 5ghz ,more importantly it has much lower memory latency, and that changes the random accesses to a large degree, also, i am polish, i know exactly what and how they tested, its just not an apples to apples comparison, its more of a overall system to system comparison, which wont translate to every scenario, if for example someone is going to use a locked slower cpu.
    But as i see you werent really interested in any valid comparison, you were just searching for something that you can bash new ryzen platform with as evidenced by all the ramblings about new ryzen platform issues true or not that have nothing to do with the actual test.
    have a nice aneurysm ;)
     
    Evildead666, Astyanax and Jackalito like this.

Share This Page