Intel DDR5 experiences

Discussion in 'Processors and motherboards Intel' started by alanm, Jan 24, 2023.

  1. alanm

    alanm Ancient Guru

    Messages:
    12,234
    Likes Received:
    4,436
    GPU:
    RTX 4080
    Was having a tough time with my first DDR5 kit (Corsair Vengeance 5600 32gb). My system would go through lengthy memory training and delayed bios posts with anything XMP or above. Although easy to OC to 6200 @ 1.3v, in aida stress tests would get a bit warm (60c+). Overall not happy with the kit due to the lengthy bios posts, so ordered a Kingston Fury Beast 5600 kit (Hynix) to test out.

    Night and day difference. From all the info I gathered so far is that Corsair DDr5 (under 6200mt kits) is mainly Samsung B-die and these can be finicky to play with. Most 5600 kits can usually do up to 6200-6400, so many of them end up being sold as 6000-6400 kits with higher voltage. The ones with low timings or above 6400 is where you will usually get the quality stuff (Hynix A-dies). Samsung vs Hynix in DDR4 may be another story, but with DDR5 this the way it seems to be.

    Anyway, from the moment I put these Hynix (M-die) sticks in, all the bios issues I had with the Corsair kit vanished. Began OC'ing them and virtually anything I tried worked. They run much cooler than the Vengeance sticks (poor heat spreaders) and got some decent results. According to buidzoid who has tested tons of DDR5 kits says Hynix is much more consistent whereas Samsung may be hit and miss. Kingston apparently does not use DDR5 Samsung memory (or very little of it) but mostly Hynix, hence the reason I went for it.

    Anyone with DDR5 observations care to share?

    5600 > 6600mt/s @ 1.38v 44-44-44-90 (1 hr stress tested)
    6600.jpg

    superseded by this with tighter sub-timings...
    7000-2.jpg
     
    Last edited: Jan 30, 2023
    bobnewels likes this.
  2. Loster

    Loster Master Guru

    Messages:
    241
    Likes Received:
    80
    GPU:
    TUF 4090
    aida0.jpg
     
    alanm likes this.
  3. bobnewels

    bobnewels Maha Guru

    Messages:
    1,413
    Likes Received:
    1,054
    GPU:
    amd
    Just to share my experience with the above mentioned DDR5 kit (Corsair Vengeance 5600 32gb)
    I have not had any problem with this Ram and Motherboard combination.

    So two motherboards with same Ram and different results,luck of the draw and in reality Ram speed does not matter for my use case 4K PC Gaming 99% of the time.I get the same FPS with Ram set 5600MT

    ♦DDR5 Ram- CORSAIR Vengeance 32GB (2x16GB) DDR5 5600MHz C36 (CMK32GX5M2B5600C36)
    ♦Mobo - ASUS Prime Z-690 P DDR5
    ♦Simple overclock on 5600Mt to 6600Mt @ 1.38v 40-40-40-76 (20 minute test since I know Ram is solid)
    CMK32GX5M2B5600C36 Set CL40 6600MT.jpg
     
    alanm likes this.
  4. alanm

    alanm Ancient Guru

    Messages:
    12,234
    Likes Received:
    4,436
    GPU:
    RTX 4080
    Cool. Nice improvement over your previous result. Mine was first attempt a few hours after receiving the kit. Fine tuning still to come. 5600 kits seem to be the sweet spot in terms of headroom potential.

    Pretty sure my issue with the corsair kit was board synergy but glad its over with. DDR5 seems to be improving in leaps and bounds, so would not be surprised if 7000mt kits become the sweet spot a year from now.
     

  5. cucaulay malkin

    cucaulay malkin Ancient Guru

    Messages:
    9,236
    Likes Received:
    5,208
    GPU:
    AD102/Navi21
    Is vccsa unlocked on the 13600 ?
     
  6. mbk1969

    mbk1969 Ancient Guru

    Messages:
    15,540
    Likes Received:
    13,556
    GPU:
    GF RTX 4070
  7. alanm

    alanm Ancient Guru

    Messages:
    12,234
    Likes Received:
    4,436
    GPU:
    RTX 4080
    Yep. Dont know why it says 13600, its a K.

    XTU.png
     
    cucaulay malkin likes this.
  8. bobnewels

    bobnewels Maha Guru

    Messages:
    1,413
    Likes Received:
    1,054
    GPU:
    amd
    The CORSAIR Vengeance 32GB works fine also on AMD system.The ram also does not need to be tuned on Intel/AMD systems and is very easily overclockable with low voltage.

    ♦DDR5 Ram- CORSAIR Vengeance 32GB (2x16GB) DDR5 5600MHz C36 (CMK32GX5M2B5600C36)
    ♦Mobo - ASRock X670E Steel Legend
    ♦Simple overclock on 5600Mt to 6200Mt @ 1.25v 36-36-36-76 (30 minute test since I know Ram is solid)
    ♦The memory clock (mclk), the memory controller clock (uclk), and the infinity fabric clock (fclk) are all synced for best results 1:1:1
    CMK32-GX5-M2-B5600-C36.jpg
     
  9. alanm

    alanm Ancient Guru

    Messages:
    12,234
    Likes Received:
    4,436
    GPU:
    RTX 4080
    @bobnewels just curious, what does the SPD tab in CPUZ say about the dram manufacturer? Is it same kit you previously showed running 6400 at over 1.4v?
     
  10. bobnewels

    bobnewels Maha Guru

    Messages:
    1,413
    Likes Received:
    1,054
    GPU:
    amd
    TGhis is the same kit for what I showed on Intel/AMD systems

    The Ram (CMK32GX5M2B5600C36) can run at default voltage @ 1.25v at higher speeds just would not be 100% stable. I am not sure why you are concerned with DRAM 1.435v I showed at 6400MT as it is not important or does not even hit High DRAM Voltage Mode.

    So on Intel/AMD at default voltage 1.25v it can run and pass stability test all day long at 6200Mt.Over 6200Mt it needs more voltage on the DRAM and that more voltage does not matter especially when it does not even go into High DRAM Voltage Mode.

    When I set the DDR5 Kit(CMK32GX5M2B5600C36) 6600Mt @ 1.38v ,I was just copying what you show for voltage.Just to show that the DDR5 Kit(CMK32GX5M2B5600C36) can actually run fine on the PC Gear I was using.

    My opinion is that the DDR5 Kit(CMK32GX5M2B5600C36) is very cheap,overclockable DDR5 that is very good in PC Gaming.
    CMK32GX5M2B5600C36.jpg
     

  11. bobnewels

    bobnewels Maha Guru

    Messages:
    1,413
    Likes Received:
    1,054
    GPU:
    amd
    Added this for fun 7000MT@1.4v
    ♦DDR5 Ram- CORSAIR Vengeance 32GB (2x16GB) DDR5 5600MHz C36 (CMK32GX5M2B5600C36)
    ddr5 (CMK32GX5M2B5600C36)7000.jpg
     
  12. alanm

    alanm Ancient Guru

    Messages:
    12,234
    Likes Received:
    4,436
    GPU:
    RTX 4080
    For now have settled on 6600 (40-40-40-84) as my 24/7 ram oc. Adjusted sub-timings for more performance. These made more of a difference than I expected vs earlier results. In DDR5 they seem to be more impactful than primary timings.

    2-6600, subtimings.jpg

    2 hour stress test:
    2-11.jpg
     
    Last edited: Jan 27, 2023
  13. alanm

    alanm Ancient Guru

    Messages:
    12,234
    Likes Received:
    4,436
    GPU:
    RTX 4080
    Ah, the lets-see-if-it-boots OC :D. Also did that and managed to run a bench. Stress test lasted about 10 secs. Seems 1.41v not enough for stability. Not going to mess further with it for time being.

    7000-2.jpg
     
    bobnewels likes this.
  14. bobnewels

    bobnewels Maha Guru

    Messages:
    1,413
    Likes Received:
    1,054
    GPU:
    amd
    You should do some test now and share them,would like to see your results.
    Tweak timings with reasonable DRAM voltage at 5600MT and whatever other speed you want and see what the results are.You may be surprised

    I would post my finding but I have posted them all over this site but here is some interesting testing from overclock.net
    DDR4 4400 C16 1T Gear1 vs DDR5 8000 C36 performance with 13900K and RTX 4090
    (2) DDR4 4400 C16 1T Gear1 vs DDR5 8000 C36 performance with 13900K and RTX 4090 | Overclock.net

    Modern Warfare 2 (2022) on High/Ultra settings without DLSS or upscaling
    8000 c36 DDR5 results [​IMG]
    4400 c16 DDR4 results [​IMG]

    Cyberpunk 2077 on Ultra with RT, DLSS on Quality, Crowd density set to High
    Results from the the built-in benchmark:

    8000 c36 DDR5 [​IMG]
    4400 c16 DDR4 [​IMG]

    Results from actual gameplay testing in which I did the exact same thing for both
    [​IMG]


    The Witcher 3 v1.32 - Maxed out
    I took the readings on my old 5900X and 3080 system for comparison before the RT update, since I was looking at these for personal use I stuck to version 1.32 for comparitive purposes.
    [​IMG]


    F1 22 - Bahrain Dry in-game benchmark, everything maxed out with relevant RT things on
    [​IMG]


    Fallout 4 - Life in the Ruins modlist from Wabbajack, it's heavily modded basically
    [​IMG]


    SkyrimSE - Elysium Remastered modlist from wabbajack that is also very heavily modded
    [​IMG]
     
  15. alanm

    alanm Ancient Guru

    Messages:
    12,234
    Likes Received:
    4,436
    GPU:
    RTX 4080
    Buildzoid has done extreme low primary timings OCs (24-29-29-28-CR1) on DDR5 but not worth the results from what I've seen, so not going to mess with it. Games not affected by mem bandwidth but favor low latency may show entirely different results, so can understand that.

    BZ's channel is great for memory OC'ers because he has tons of varied examples of DDR4-5 testing on Asus, MSI and Gigabyte motherboards with Samsung, Hynix memory. Learned a lot from it.


     
    Maddness likes this.

Share This Page