Does Single Channel DDR5 hurt performance?

Поділитися
Вставка
  • Опубліковано 5 лип 2024
  • We know that running a single stick of DDR4 harms performance, but DDR5 is built different: Each stick has 2 channels. So we set about testing a single stick to see if there's a performance difference. Can the new design mitigate some of the detriment of running just one stick?
    ______________________________________________________________
    Timestamps:
    0:00 - Introduction
    0:31 - DDR5 is built differently
    1:28 - Cinebench R23
    1:51 - 3DMark Time Spy
    2:07 - Shadow of the Tomb Raider
    2:45 - Forza Horizon 5
    3:04 - Flight Simulator 2020
    3:18 - Result Conclusions
    3:56 - A 64 bit bus is still just a 64 bit bus
    4:25 - Overall Recommendations: Stick with 2 sticks so you don't get stuck with one stick.
    ______________________________________________________________
  • Наука та технологія

КОМЕНТАРІ • 57

  • @volkingdeath5312
    @volkingdeath5312 2 роки тому +32

    This channel deserves way more views for such high quality videos

    • @treyquattro
      @treyquattro 2 роки тому +1

      it'll get them - quality always rises. It can just take a frustrating amount of time with YT. Always "like" and spread the word!

    • @PrashantMishra-kh1xt
      @PrashantMishra-kh1xt 11 місяців тому

      Too bad he hasn't uploaded in 8 months

  • @ankitlal5931
    @ankitlal5931 Рік тому +3

    Exactly what I was looking for, great video

  • @kenobi639
    @kenobi639 Рік тому +1

    Great channel. Great presentation. Great information. Criminally under-subscribed

  • @ChristopherHailey
    @ChristopherHailey 2 роки тому +2

    Lots of useful info here, thanks

  • @RAM_PRATHEESH
    @RAM_PRATHEESH 2 роки тому +1

    Nice video man

  • @hongockimquang1994
    @hongockimquang1994 2 роки тому +1

    I clicked. I liked. I watched. Such simple chores only available on good straight-to-the-point videos.

  • @eliasali9383
    @eliasali9383 Рік тому +1

    That's just what I needed
    thank you

  • @Bugatti12563
    @Bugatti12563 Рік тому +5

    starting my new build with single 16 gb stick of ddr5 ram. I'm just glad that even works!

  • @VijayRana98554
    @VijayRana98554 6 місяців тому

    thanks for providing this valuable information for us😊

  • @bindaasmard664
    @bindaasmard664 Рік тому +1

    Excellent video, I was curious how would single 32gb 6000 CL30 stick would perform against 2*16gb CL36 kit ? in my country the newly launched XPG (AX5U6000C3032G-CLARBK) CL30 32gb stick costs way too less when compared with 2*16gb CL36 kit.

  • @gorbokonyk
    @gorbokonyk Рік тому

    100% useful. Txns!

  • @R1b3z
    @R1b3z Рік тому +5

    I had to go with one stick because it was noticeably cheaper then 2x8gb and I can later buy another of 16 and get dual channel at 32gb, I don’t think it will be hard to find a matching hyperx stick because where I live they are the most common ddr5 dimms so yah, now I just wait for a sale or something and get another one, performance is still great, and since I wanted to OC my i5-12400f in my country I could only find ddr5 boards with external bclk clock so I had to change to ddr5 and just sell the old board with the ddr4 sticks, performance is amazing but is good to know I will get more once I go to dual channel again :)

  • @gasiedu
    @gasiedu Рік тому +1

    damn would of been good if i ran into this video before i bought my ram, great vid

  • @enescetiner390
    @enescetiner390 Рік тому

    would you prefer kingston fury beast 2x8 6000mhz CL36 or 1x16 5600mhz CL36 for a bit cheaper?

  • @WhirlwindHeatAndFlash
    @WhirlwindHeatAndFlash 2 роки тому

    very nice data

  • @yooanto9465
    @yooanto9465 10 місяців тому

    Is there any test on C latency for DDR5 look at 99% fps low. It can fix slight stuttering issue many reports to have on New intel 12th 13th cpu.

  • @carptackula7536
    @carptackula7536 2 роки тому +2

    This was an AWSEOME comparison - that you sooo much for this benchmark! I have a question not addressed: DDR5 is (now) available in (2x8GB) would it be better to go with this vs 1x16gb? (It sounds like 'it would' - ie 'dual channel is always better')

    • @BuildPicker
      @BuildPicker  2 роки тому +1

      YEs, dual channel is always better, DDR4 or DDR5! Doubles your bandwidth to the memory.

    • @terry-
      @terry- Рік тому +1

      @@BuildPicker altho, looking at the numbers, it seems like the same thing. more or less 10% for dual rank and another 10% for dual channel. It would be very interesting to see this theoreticall 20% by testing 1x8gb vs 2x8gb vs 1x16gb vs 4x8gb vs 2x16gb vs 4x16gb.

  • @KetogenicGuitars
    @KetogenicGuitars Рік тому +4

    Absolutely you can. That ~10% means nothing(for me). It is good update path if there is prospect prices coming down in like 6-12 months. There was a test about matching unrelated sticks by Linus Sebastian and they found it surprisingly unmeaningful thing. But how about choosing one of the most common sticks like Viper Steel 6000Mhz 32GB?

  • @Takashita_Sukakoki
    @Takashita_Sukakoki 2 роки тому +2

    I would be interested how 2x8gb ddr5 compares to 2x16gb kit with rest equal couldnt find any real data on it. Most people dont need 32gb but how big could the single sided vs dual sided (dual channel config) stick impact be?

    • @BuildPicker
      @BuildPicker  2 роки тому +1

      I've only seen 2x8GB kits available in the very lowest speeds/specs, which I wouldn't recommend anyway. If 2x8GB DDR5 is 4800MT/s and $126... just get 2x16 GB DDR4 @ 3600Mhz. It'll perform better and be the same price.

    • @Takashita_Sukakoki
      @Takashita_Sukakoki 2 роки тому +1

      There are ddr5 2x8gb kits from 4800-6000 from kingston so the question still stands do you NEED to grab a 32gb kit for new platforms that no longer support ddr4.

  •  2 роки тому +1

    It was a question that I could not find the answer to for a long time. Thank you for clarifying.

  • @Ace-mr8cc
    @Ace-mr8cc 7 місяців тому

    Not sure if its just me but I just purchased a New PC and it came with a single 16gb DDR5 Ram stick and I'm experiencing a ton of stuttering/FPS Spikes with over 100 FPS..

  • @dalexa
    @dalexa 2 роки тому +2

    I might have missed this so I apologise. Is the ddr5 always the same size, i.e., 2x16 vs 1x32 (or similar) or is it 1x16 vs 2x16?

    • @BuildPicker
      @BuildPicker  2 роки тому +3

      1x16 vs 2x16, as we don't have any 32GB Dimms to test at this time!

  • @covfefe9661
    @covfefe9661 5 місяців тому

    informative😊

  • @priyanshkumar308
    @priyanshkumar308 6 місяців тому

    The voice is little problematic to understand
    Maybe editing the voice would be better but the content was very helpful👍
    Thankyou

  • @stevencen4076
    @stevencen4076 2 роки тому +1

    I regret didnt see this video before buying my pc part. I on progress to build my first pc so im still new to this things and i just bought asus rog strix b660 f that can only use ddr 5 and i miscalculate the entire budget because i dont know ddr 5 ram will be that expensive and the performance is still below ddr 4 :(

  • @treyquattro
    @treyquattro 2 роки тому +3

    another excellent review with pertinent results. Besides trying to decide between Z690 and B660 right now, the other decision is between DDR4 and DDR5, and I'm struggling to justify DDR5 given the cost and no appreciable performance gain (depending on workload mix - sustained transfers vs. latency-inducing disparate transfers. e.g.). I do expect DDR5 performance to improve, but JEDEC apparently says this will be in increments of 400Mbps per year*, so seemingly there isn't a big rush. The question is whether it's worth the short term financial pain to throw the dice on future-proofing. With PCIe 5.0 and Raptor Lake apparently fitting in an LGA1700 socket, the answer just might be Yes.
    * I got that figure from an interview with a Kingston exec on Igor's Lab.

    • @carptackula7536
      @carptackula7536 2 роки тому

      DDR5 is faster then DDR 4 in a few different situations: Audio Production (VSTs) and Adobe Premiere. Everything else DDR4 3600 (low latency) is equivalent and half the price, so you can buy twice as much

    • @treyquattro
      @treyquattro 2 роки тому

      @@carptackula7536 I went with DDR5 for my Alder Lake build. The memory was expensive but I'm satisfied with it. I intend to upgrade when DDR5 approaches current DDR4 prices. However, I have 64GB so that will suffice for a while. You can't wait forever - there will always be newer, better tech, and the older stuff will get cheaper.

  • @marwanalkhaleel4176
    @marwanalkhaleel4176 2 роки тому +1

    What if we have 3 DDR5 sticks, is that going to disable dual/quad channel when the 3rd stick is added?

    • @BuildPicker
      @BuildPicker  2 роки тому +1

      Would normally run flex mode, so dual channel for the first 32GB, and single channel when the OS accesses that last 16GB.

    • @marwanalkhaleel4176
      @marwanalkhaleel4176 2 роки тому

      @@BuildPicker Thanks for the quick reply. We have two 16GB DDR5 sticks, so 32GB total. And thinking of adding more ram and wondering if we should go for another two 16GB sticks or a single 32GB stick. Any advice? If I'm not mistaken if there's any performance dip it'll be less than 10%. Or maybe a clearer question, since the 3rd stick is equal to the other two's capacity will they run as if they're in quad channel or is that too much of a long shot? lol

    • @terry-
      @terry- Рік тому

      @@marwanalkhaleel4176 if you already have a dual channel memory config working, add another kit of 2 sticks to add size and keep working on dual channel. but using 4 sticks of ram, depending on the motherboard could limit the final ram speed that you can ran at. meaning, 4800 instead of 5200 for the 4 sticks, for example.

  • @metromelvin
    @metromelvin 2 роки тому +1

    Nice,, :)

  • @Ortagonation
    @Ortagonation 2 роки тому

    we cannot test it until am5 coming this summer tho' (or new intel)

  • @Micky999
    @Micky999 Рік тому

    So for 1080p/60fps gaming, going to 1 x DDR5 6000MT/s is this OK? notice i don´t need more than 60fps.

    • @BuildPicker
      @BuildPicker  Рік тому +2

      I'd be building a ddr4 system. Far better value.

    • @Micky999
      @Micky999 Рік тому

      @@BuildPicker thank you.

  • @raghupathy9396
    @raghupathy9396 Рік тому +1

    What would be the difference bw 2x8 ddr4 vs 1x16 ddr5 ?

  • @WholesomePotato1
    @WholesomePotato1 11 місяців тому

    Thanks for this video. The person who built my computer clearly doesn't know what he's doing. I won't mention everything he did wrong because that would be a side-track, but this one honestly pissed me off. He put a single stick of 32 GB DDR5 6000Mhz RAM and effectively reduced performance of my computer by 10-20%. My other stick should arrive tomorrow.
    I have just 2 questions for you since you clearly know what you're talking about - maybe a silly question, but bear with me - I ordered an RGB version of the same RAM stick but I have non-RGB in my PC now. Is that OK?
    And secondly, if I have a RAM stick in the first slot, the second stick should go to 3rd? I have ASUS Prime B650 Plus motherboard.
    Thank you and have a great day ^.^

  • @bodaciouschad
    @bodaciouschad 2 роки тому +2

    2:38 How accurate can this data be? 4000mhz c15 was slower than 3600mhz c16 in your testing. Surely that can't be true on average across many repeated tests, as the 4000mhz c15 kit is both faster and lower latency.

    • @BuildPicker
      @BuildPicker  2 роки тому +7

      Yes, I explain why elsewhere. The 4000MHz Cl15 results are overclocked and tweaked, but not optimised across all timings/sub timings. Depending on the nature of the test, sometimes it wins against 3600MHz Cl16 at XMP, and sometimes it loses. XMP has a whole load of timings and sub timings that are applied, included in BIOS microcode, that are not user accessible. The B-Die DDR4 results are intended as an example of tuned B-Die performance, not a definitive assessment of it. It's the same with 4400Mhz Cl19 in gear 2 - Bandwidth heavy tests favor that, Latency dependent tests penalise it.
      That result is also the total average FPS, so the render thread, GPU etc comes into the mix as well. The 'game engine' results show much more granularity in the results between RAM, but also aren't so relevant to the real world, I.e. the performance that you actually obtain when everything else is brought into the mix. One of the reasons we use the Shadow of the tomb raider benchmark is because of it's repeatabliity. We can re-build systems and apply the same parameters, and you get the same results. It's remarkably consistent.
      I hope that helps explain. There's many factors at play here, and whilst we do try and control them as best we can, real world data is sometimes messier than we'd like.

  • @wildwoodtop
    @wildwoodtop Рік тому

    32GB DDR5 4400MHz RAM would this be two sticks or one 🤔

    • @BuildPicker
      @BuildPicker  Рік тому

      Normally 2 - DDR5 Kits are generally 16GB per stick but 8GB kits do exist now.

  • @martintomas6537
    @martintomas6537 2 роки тому

    You are me hero

  • @scheme4685
    @scheme4685 Рік тому +1

    How does 4 sticks compare to 2 sticks in DDR5?

    • @BuildPicker
      @BuildPicker  Рік тому

      Much harder to stabilise, so it will generally be slower.

  • @davidvannfa440
    @davidvannfa440 Рік тому

    well, i was interested to see the 4k performance, i guess i'll have to look elsewhere

    • @BuildPicker
      @BuildPicker  Рік тому +1

      There will be no difference as you're GPU limited then anyway. That's why GPU and RAM benchmarks are run at 1080p or even lower.

  • @grampsradio
    @grampsradio 8 місяців тому

    monotone