*Official* NBR Desktop Overclocker's Lounge [laptop owners welcome, too]

Discussion in 'Desktop Hardware' started by Mr. Fox, Nov 5, 2017.

  1. electrosoft

    electrosoft Perpetualist Matrixist

    Reputations:
    2,719
    Messages:
    3,937
    Likes Received:
    3,382
    Trophy Points:
    331
    That must be such a nice luxury....nice. :)
     
    Clamibot, ajc9988, Rage Set and 2 others like this.
  2. Johnksss

    Johnksss .

    Reputations:
    11,523
    Messages:
    19,409
    Likes Received:
    12,639
    Trophy Points:
    931
    In reality, yes! Wish is was like that for EVGA, since they are just as close.
     
  3. ajc9988

    ajc9988 Death by a thousand paper cuts

    Reputations:
    1,743
    Messages:
    6,111
    Likes Received:
    8,823
    Trophy Points:
    681
    upload_2021-2-18_12-18-49.png

    So, I cannot get it much tighter than this with these specific DIMMs. One note, if you cannot get the write to read delay to go down by setting tWTR_L and tWTR_S, use tWRRD_sg and tWRRD_dg. Lowering those lowers it on the IMC side for the ram timings.

    I tried to get the board to train with lower RTL and IO-L. With Gigabyte, the only way to force it to use your settings in the BIOS is to, after setting it in BIOS, turn off the power entirely. When you turn it back on, it will try to train the ram. I could never get it to train with the tighter RTL and IO-L, but that is for anyone else running a Gigabyte board.

    I couldn't get CL15 and CWL12 to fully train stable. Maybe if I left the write to read delay looser, or the tRFC. But the performance is not drastically difference. Still have more testing, but Karhu and HCI Memtest stable is pretty stable. Just have some more to go. Remember (more for lurkers reading or those less familiar with the process of testing memory), you must test not only that the memory is stable, but that the memory WITH the cache included in stressing is stable. So, for the programs that allow you to include the cache (although testing with cache and passing is prima facie it is good), you should run it both with and without the cache.
     
  4. Johnksss

    Johnksss .

    Reputations:
    11,523
    Messages:
    19,409
    Likes Received:
    12,639
    Trophy Points:
    931
    What happens if you raise the tREFI?
     
  5. ajc9988

    ajc9988 Death by a thousand paper cuts

    Reputations:
    1,743
    Messages:
    6,111
    Likes Received:
    8,823
    Trophy Points:
    681
    tREFI can help, generally, but has a diminishing returns on how it helps with performance. tREFIx9 should be maxed out if possible. It is the window in which however many tREFI occur. For tREFI, people often max it out, and it can seem stable, but it is harder to test if it is actually stable. One suggestion for testing tREFI because it is the refresh interval is to get a large file and put it into memory (like by creating a drive) and hold it there for a period of time long enough that errors can accrue due to information loss from not refreshing soon enough causing data errors. You then do a checksum of some type to compare the hash of the original file and the later file.

    This will not usually effect general files or use in memory, so most recommend to take whatever the auto sets, up it by 5,000 at a time until you see no increase in performance, then stop.

    The refresh interval is when the charge in the cells is refreshed. If the cells lose their charge before the content of the cell is copied, you can lose the information or wind up with corruption. You will usually see this problem from other timings before you see it from setting the tREFI, but there is always that chance. If you don't get a performance lift by increasing the amount of time before it refreshes the cells, then no reason to have it higher than it needs to be. If you have the refresh interval too low, then the memory has to copy the information out to refresh the cells more often than is necessary, thereby lowering the performance. So higher is better, until you see no more gains, then stop to prevent data corruption.
     
    Papusan, Rage Set and Johnksss like this.
  6. Johnksss

    Johnksss .

    Reputations:
    11,523
    Messages:
    19,409
    Likes Received:
    12,639
    Trophy Points:
    931
    So in other words, it will help, but comes with a price if you over use it. Got it.:)
     
    Papusan, Rage Set and ajc9988 like this.
  7. ajc9988

    ajc9988 Death by a thousand paper cuts

    Reputations:
    1,743
    Messages:
    6,111
    Likes Received:
    8,823
    Trophy Points:
    681
    Although I disagree with some aspects of what are written here, this guide seems good enough to share. It hits the timings in a different order than I do, so that may be able to give a better result. I'll probably try this guide to OC my ram just to see if doing it this way yields any better performance than I already have doing it my way (always open to try something new; if it works, then cool, if not, then I just wasted some time).

    https://github.com/integralfx/MemTestHelper/blob/master/DDR4 OC Guide.md/#timings-software
     
    Rage Set and Papusan like this.
  8. electrosoft

    electrosoft Perpetualist Matrixist

    Reputations:
    2,719
    Messages:
    3,937
    Likes Received:
    3,382
    Trophy Points:
    331
    Well, it looks like I'm keeping the 3090 (my father always said never mix money and friends....and he was right again), so I listed my 3070 on eBay for $1k (yes, a thousand) making it slightly cheaper by ~$100-150 than the last 5-10 on average that had sold and it sold within an hour.

    This cryptocraze is out of control.
     
    Rage Set and ajc9988 like this.
  9. ajc9988

    ajc9988 Death by a thousand paper cuts

    Reputations:
    1,743
    Messages:
    6,111
    Likes Received:
    8,823
    Trophy Points:
    681
    I wish it was just a crypto craze. When it is
    effecting auto production due to chip shortages, the problems are far beyond simply crypto miners.


    In other news, my manifold arrived.
    [​IMG][​IMG][​IMG]

    Edit: each of those is a quick disconnect with the rubber boot for protection still on it until use.
     
    Johnksss, Rage Set and electrosoft like this.
  10. Papusan

    Papusan Jokebook's Sucks! Dont waste your $$$ on Filthy

    Reputations:
    41,655
    Messages:
    29,544
    Likes Received:
    58,538
    Trophy Points:
    931
    And here is some more problems Red camp struggle with.

    AMD confirms USB problems with 500 series motherboards computerbase.de
    The increasing number of users who complain of problems with USB ports on motherboards with B550 and X570 chipsets calls AMD onto the floor.

    AMD has now taken a closer look at the malfunctions relating to USB connections on motherboards with AMD chipsets of the 500 series, which have been reported by several users for weeks , and is using the growing number of reports as an opportunity to get to the bottom of the problem. A solution is currently not in sight.

    AMD is aware of reports that a small number of users are experiencing intermittent USB connectivity issues reported on 500 Series chipsets. We have been analyzing the root cause and at this time, we would like to request the community's assistance with a small selection of additional hardware configurations. Over the next few days, some r / Amd users may be contacted directly by an AMD representative (u / AMDOfficial) via Reddit's PM system with a request for more information.
     
    iunlock, Johnksss, Rage Set and 2 others like this.
Loading...

Share This Page