Not happy with the new bios from MSI...

Discussion in 'Processors and motherboards AMD' started by waltc3, Apr 12, 2018.

  1. waltc3

    waltc3 Maha Guru

    Messages:
    1,445
    Likes Received:
    562
    GPU:
    AMD 50th Ann 5700XT
    I installed 2.A0 (replacing 2.8 in my Gaming Pro Carbon AC) and my AIDA64 cpu bench numbers have all dropped by 10% or more..! Don't tell me this is because of that Spectre 2 garbage...! Aggghhh...not worth it especially as there are no Spectre 2 exploits known in the wild, anywhere right now. I'm going to look at a few more things to make sure I am not hallucinating, but will likely be going back to 2.8 if I don't find I've made a mistake somewhere! Bios was just released today, apparently, as there was no sign of it yesterday! Very poor, if this is correct.
     
  2. waltc3

    waltc3 Maha Guru

    Messages:
    1,445
    Likes Received:
    562
    GPU:
    AMD 50th Ann 5700XT
    Looks like the slowdown is real--from 7%-12%, anyway. MSI had the bright idea of making it impossible to reflash an back to older version--renaming and re-*SPAM* the file didn't help. This is the first fairly upsetting experience I've had so far with this motherboard. Coincidence that that the bios release coincides with AMD's Spectre 2 press release, which talked about bios fixes sent to the board OEMs? Wouldn't think so. Great. BTW, I used the same version of AIDA 64--and I had saved my previous runs so that they'd show up--slowdown seems real at this point. If anyone has any ideas please respond. I thought it might be even the Win10x64 cumulative update I had installed--so I uninstalled that--but it wasn't. Pretty soon this security mania of releasing performance-damping "fixes" without there being any malware to prompt it is going to have us all locked down to where we can't move... prison is a secure place to be I would imagine, but who wants to live there?...;)
     
  3. user1

    user1 Ancient Guru

    Messages:
    2,746
    Likes Received:
    1,279
    GPU:
    Mi25/IGP
    you can try using the inspectre tool, to disable protection and see if it is the cause, i think the option to disable it is also somewhere security settings if you dont want to use it
     
  4. RealNC

    RealNC Ancient Guru

    Messages:
    4,893
    Likes Received:
    3,168
    GPU:
    RTX 4070 Ti Super
    CPU benches are one thing. You should instead bench your actual workloads. There should be a registry key that controls Spectre mitigation (the microcode doesn't do anything on its own; the OS is what uses it to implement the mitigation methods.)

    Bench your actual workloads with the mitigations enabled and disabled.
     

  5. waltc3

    waltc3 Maha Guru

    Messages:
    1,445
    Likes Received:
    562
    GPU:
    AMD 50th Ann 5700XT
    Thanks for the suggestions, guys--will investigate further and get back here with the results!
     
  6. Kool64

    Kool64 Ancient Guru

    Messages:
    1,655
    Likes Received:
    782
    GPU:
    Gigabyte 4070
    It dropped the performance of my X370 Gaming Plus as well. There were more issues than I could count. I was able to force a BIOS downgrade to go back to my most stable build since AMD blocked downgrades after the 1.0.0.1 AGESA. I won't be updating my BIOS at all going forward. gainrecording in this thread https://forum-en.msi.com/index.php?topic=302340.0 has the fix for you to be able to downgrade your BIOS. It's a use at your own risk type thing but the process worked great for me.
     
  7. waltc3

    waltc3 Maha Guru

    Messages:
    1,445
    Likes Received:
    562
    GPU:
    AMD 50th Ann 5700XT
    Thanks--I am really curious about this. I tried--could never get it to work from my USB Thumb drive--kept giving me the following error messages:

    Unable to find EFI Update script
    Please mount the drive with the update script

    And I couldn't even get that error message until after I had turned off secure boot--else I could never get past that point. Something's missing or I'm doing something wrong, although I followed the instructions to the letter--then tried every variable I could think of. I can use that drive to do a RUFUS Win10 install--but for some reason it's not working here.

    Here's a bit of the weirdness--not sure what's going on.

    AIDA 64 5.97.4600

    R5 1600 MSI Gaming Pro Carbon AC

    Hash cpu benchmark with the 2A0 bios:

    3.2GHz 14442 MB/s ( Auto--Core performance boost turned off)

    3.4GHz 15356 MB/s (Auto--Core Performance boost ON-default)

    3.6GHz 14641 MB/s (Core boost disabled)

    3.8GHz 14749 MB/s (Core boost disabled)

    3.9GHz 14797 MB/s (Core boost disabled)

    WTF?

    By way of example, with the 2.8 Bios, @ 3.8GHz my score in the same bench was: ~17890...! Repeatedly.

    It certainly looks like some throttling is going on even though I have "Relaxed EDI throttling" ENABLED in the bios--DISABLED, makes no difference. Something in this bios is artificially limiting the performance of the cpu. Heat rarely exceeds 55C--so that's not a problem at all. Very odd. Either a bug or deliberate--I cannot say at this point. Really wish I had left well enough alone with 2.8...!
     
  8. stormy

    stormy Ancient Guru

    Messages:
    2,671
    Likes Received:
    56
    GPU:
    Pulse RX5600XT
    Walt you can now downgrade the bios if you are so inclined. Svet on the MSI forums posted this,

    https://forum-en.msi.com/index.php?topic=302638.0

    two days ago. I just used it on my B350 Tomahawk to go back to the last bios before the AGESA Code 1.1.0.1 update.Worked without a hitch, and fixed a game issue I had been having (one game only, and something in the latest bios that it really didn't like).
     
  9. RealNC

    RealNC Ancient Guru

    Messages:
    4,893
    Likes Received:
    3,168
    GPU:
    RTX 4070 Ti Super
    What's the point of downgrading the BIOS if Windows Update will soon install the microcode anyway?
     
  10. Agent-A01

    Agent-A01 Ancient Guru

    Messages:
    11,628
    Likes Received:
    1,119
    GPU:
    4090 FE H20
    It's easy to undo it, just rename microcode update file.
     

  11. RealNC

    RealNC Ancient Guru

    Messages:
    4,893
    Likes Received:
    3,168
    GPU:
    RTX 4070 Ti Super
    And what if the new microcode has other, unrelated fixes in it? Like performance improvements or bug fixes; these are usually silent and no one reports about them.

    To me it looks like you should keep the latest microcode and just disable the mitigations if you don't want them.
     
  12. Kool64

    Kool64 Ancient Guru

    Messages:
    1,655
    Likes Received:
    782
    GPU:
    Gigabyte 4070
    This latest round of BIOS updates for AMD boards is totally messed up. I guess a lot of it is the AGESA rewrite. I updated mine because it claimed improved memory compatibility which I’m always looking for. Needless to say it was not helpful to me and it broke quite a few things. For instance I could not get a stable OC past 3.8(was at 3.925) so I decided just to run stock for a while to wait for them to sort out their issues but the auto voltage was giving the chip 1.46 volts and higher.

    Also another strange issue was that when OC’ing the CPU as speed increased the memory bandwidth would actually go down with the same memory speed and timings. It was a train wreck of an update to say the least.
     
  13. waltc3

    waltc3 Maha Guru

    Messages:
    1,445
    Likes Received:
    562
    GPU:
    AMD 50th Ann 5700XT
    I wish...;) I renamed it and I used a program to change the modified file date--still prevented the downflash. Must be something with in the file code that it uses.
     
  14. waltc3

    waltc3 Maha Guru

    Messages:
    1,445
    Likes Received:
    562
    GPU:
    AMD 50th Ann 5700XT
    Thanks much--will check it out!...;) Svet is usually right about this stuff!
    It's not the mitigations--I neglected to mention earlier that removing the Spectre 2 Windows code (Ryzen isn't vulnerable to Meltdown--so it has no Windows nor microcode applied for it) was not the problem. Since my first posts here I have seen gobs of weirdness from this bios--basically the faster I clock the cpu the slower it runs...!....;) I've never actually seen a bios version this borked...;) I'm afraid this one needs to go back to the drawing board--the previous bios was very nice. Hopefully I can get back there today.
     
  15. waltc3

    waltc3 Maha Guru

    Messages:
    1,445
    Likes Received:
    562
    GPU:
    AMD 50th Ann 5700XT
    OK...all's well, the instructions here, https://forum-en.msi.com/index.php?topic=302638.0 , worked like a charm first time, no problems. Now back on 2.8, and bench scores are again behaving as they are expected to behave--my former 17,9xx MB/s HASH bench scores @ 3.8GHz have returned! I must admit this is the first time I've ever flashed to a new bios that was demonstrably inferior to the last! Lesson learned--I will be looking before I leap in this regard from now on.

    I didn't mention it earlier--but in addition to the other problems--the 2A0 bios was booting me up at times to 2.8GHz...! That's 400MHz below default clock....;) Weird, indeed...! MSI should scrap it--pull it--and deep six it. Thanks much for helping me get this figured out and corrected guys...!
     

  16. Kool64

    Kool64 Ancient Guru

    Messages:
    1,655
    Likes Received:
    782
    GPU:
    Gigabyte 4070
    I wonder if any of this stuff is going to carry over to the 2k series chips. The issues I had with up with would render one nearly useless.
     
  17. waltc3

    waltc3 Maha Guru

    Messages:
    1,445
    Likes Received:
    562
    GPU:
    AMD 50th Ann 5700XT
    OK, I have to eat my words...it would appear that bios version 2.A0 for my motherboard was borked by the AMD microcode Spectre 2 update, and that my first instincts were correct. After I had installed 2.A0 I ran InSpectre which verified that I had all the Spectre protections enabled, including the Microcode updates. So I turned off The OS Spectre 2 protections and rebooted to no change in performance or any of the other bugs, so I concluded that it wasn't the OS mitigation--erroneously, it turns out. I had thought the microcode had been applied earlier--I was wrong. After successfully restoring the system to the very nice and stable 2.8 bios, I fired up Inspectre today--and guess what? Now it informs me that the microcode "fix" for Spectre 2 is missing (even though all of the OS stuff is there.)....! So to my mind--that's it. Mystery solv-ed, as Clouseau might say! For now, anyway, my advice is stay away, far away from the latest Ryzen UEFI bios versions! until a later version fixes it! So thoughtful of MSI not to include this information in the bios update info descriptions, eh? As well, so thoughtful of them not to make a ready escape out of it as they *had* to have known it was bad before they released it. Something this bad is so very obvious immediately--probably didn't test it--which beggars comprehension, frankly. They *knew* the code contained the Spectre 2 cpu microcode. Very sloppy and reckless, imo.
     
  18. Kool64

    Kool64 Ancient Guru

    Messages:
    1,655
    Likes Received:
    782
    GPU:
    Gigabyte 4070
    Yeah so I think their crappy BIOS cooked My1600X. For the month or so it took me to figure out how to downgrade my board was pushing 1.46V into into the CPU I could not get at stable OC so I just set everything to auto but I should have looked at the voltage earlier ;/. Now I can't OC for crap. I wonder if it's the chip or the board?
     
  19. waltc3

    waltc3 Maha Guru

    Messages:
    1,445
    Likes Received:
    562
    GPU:
    AMD 50th Ann 5700XT
    The microcode is the culprit, unfortunately, at least in my case. The OS mitigations are pretty painless. It's odd, really, how OEM common sense (at AMD and Intel both) seems on holiday with these issues. What's the point of the microcode updates if the cure is much worse than the disease? I can laugh about it, now that I have a properly functioning box again--but so far it seems the microcode "cure" is far more dangerous that the theoretical Spectre 2 code in the wild that seems so elusive when it comes to producing working code that is something other than a purely theoretical "proof" of concept... Just goes to show why, imo, these issues should never be treated with the wild, witless panic that initially was the case with these vulnerabilities. These events do much to erode the public's confidence in the efficacy of the hardware IHVs--this has certainly made an impression on me, no question about it.
     
  20. waltc3

    waltc3 Maha Guru

    Messages:
    1,445
    Likes Received:
    562
    GPU:
    AMD 50th Ann 5700XT
    Well, the Windows mitigation code supplied by Microsoft is entirely separate from the cpu microcode that AMD parcels out to the mboard OEMs to be used within an upcoming bios release. It seems that to completely mitigate Spectre 2 on Ryzen both a Windows (software) patch and a bios-delivered cpu microcode update are required. To have one installed but not the other is to have failed to address the Spectre 2 vulnerability entirely. From the InSpectre #8 readme:


    This system's present situation:

    This 64-bit version of Windows has been updated for full awareness of both the Spectre and the Meltdown vulnerabilities. If the system's hardware (see below) has also been updated, this system will not be vulnerable to these attacks.

    This system's AMD processor is not affected by the Meltdown vulnerability, but it has not been updated with the new features required to allow its operating system to eliminate the Spectre vulnerabilities and/or to minimize their impact upon the system's performance.
    ________________________________________________________

    InSpectre #8

    That last sentence reads like the above under the 2.8 bios, which does not install the Spectre 2 cpu microcode. When the Spectre 2 microcode is installed via the 2.A0 bios installation, the wording in the InSpectre detection program changes to state that the system is now fully protected from a Spectre 2 attack by the required Windows software code and the required cpu microcode being present in the system.

    I went back and looked again at my saved AIDA 64 cpu benchmark results under 2.A0 and compared them with the 2.8 results and in some cases performance drops by as much as 15%-20%...! Memory bandwidth for read, for instance, in the AIDA 64 bench, dropped from ~48408 MB/s to ~40128 MB/s--there was a similar loss in speed for memory writes. As these results always repeat, I don't think there is any question but that the Spectre 2 microcode "fix" negatively impacts the performance of Ryzen's out-of-order speculation capability, and not by a little bit, either. So I will stay on 2.8 for the time being--and hope that they'll manage to mitigate their mitigation!

    But out-of-order speculation performance seems to be the only thing hit by the Spectre 2 microcode, as oddly enough, the cpu benchmark results for Cinebench 15.xx. do not seem affected at all, and are the same with or without the Spectre 2 microcode having been applied to my R5 1600.
     

Share This Page