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

Build Advice Upgrade time?

Discussion in 'Hardware' started by Gareth Halfacree, 15 Mar 2019.

  1. Vault-Tec

    Vault-Tec Green Plastic Watering Can

    Joined:
    30 Aug 2015
    Posts:
    14,973
    Likes Received:
    3,735
    May I suggest a happy medium?

    https://www.amazon.co.uk/Intel-M-2-...ntel+1tb+nvme&qid=1552740338&s=gateway&sr=8-1

    I have one, and I can tell the difference at boot. It's £60 cheaper though and is a proper NVME.

    https://www.intel.com/content/www/u...60p-series/660p-series-1-tb-m-2-80mm-3d2.html

    The speeds on Amazon are a typo

    I know, but unfortunately there are only about three review sites on the net where PSUs are actually tested properly. Thus, many don't know that because they don't understand how they work.
     
    Last edited: 16 Mar 2019
  2. Anfield

    Anfield Multimodder

    Joined:
    15 Jan 2010
    Posts:
    7,062
    Likes Received:
    970
    FTFY

    Every non titanium spec meanwhile is just "do whatever you want" below 20%.

    But either way, the price premium isn't really worth it (well, except that sweet deal in the marketplace, even if it is total overkill as I have a T2 1KW powering a 2950x and 2080TI without breaking a sweat)
     
  3. Corky42

    Corky42 Where's walle?

    Joined:
    30 Oct 2012
    Posts:
    9,648
    Likes Received:
    388
    Agree on the titanium PSU's not being worth it, i bought a platinum and regret spending so much on it as to recuperate the £80-90 increase in price over a gold or silver would take me 10-20 years based on this calculator.
     
  4. Gareth Halfacree

    Gareth Halfacree WIIGII! Lover of bit-tech Administrator Super Moderator Moderator

    Joined:
    4 Dec 2007
    Posts:
    17,132
    Likes Received:
    6,728
    Just to add to the "why do you need all the cores" explanation:

    upload_2019-3-19_9-38-40.png

    I've been waiting for this thing to finish on my poor ol' A10-5800K for about fifteen minutes now, and I've got a fair while longer to wait yet. If I had eight processor cores, instead of somewhere between two and four depending on whether you buy AMD's Bulldozer-era definition of a 'core' or not, I'd be done in a quarter of the time - less when you figure that said cores will also be faster, as well as more numerous.

    For now, though, I'mma switch to doing something else while this gets on with things in the background.
     
  5. Gareth Halfacree

    Gareth Halfacree WIIGII! Lover of bit-tech Administrator Super Moderator Moderator

    Joined:
    4 Dec 2007
    Posts:
    17,132
    Likes Received:
    6,728
    Sorry, missed this earlier: that's certainly cheaper, but it's also a lot slower because it's QLC NAND. Check out the figures here - it benchmarks (which, yes, is not fully indicative of real-world performance) at nearly a third the data rate of the Samsung.
     
  6. yuusou

    yuusou Multimodder

    Joined:
    5 Nov 2006
    Posts:
    2,878
    Likes Received:
    955
    QLC performance really starts to tank after 50%~ full.
     
  7. Gareth Halfacree

    Gareth Halfacree WIIGII! Lover of bit-tech Administrator Super Moderator Moderator

    Joined:
    4 Dec 2007
    Posts:
    17,132
    Likes Received:
    6,728
    Aye, so instead of buying a £117 1TB QLC which drops to sub-SATA speeds when you use more than 500GB, I could buy a 500GB 970 Evo for the same price and get full performance from 0-100% usage. Or just pay the extra for the 1TB version, which I reckon is worth it.
     
  8. bawjaws

    bawjaws Multimodder

    Joined:
    5 Dec 2010
    Posts:
    4,284
    Likes Received:
    891
    Just to revisit this quickly: I have a 500GB Samsung 960 EVO (NVMe) and a 500GB 850 EVO (SATA, 2.5") in this PC and although benchmarks show that the former has read and write speeds of about five and four times the latter respectively, when I've switched applications and games between the two drives I have noticed absolutely no perceptible difference between the two drives. The desktop is just as snappy with the OS installed on either drive, too.

    Now not all SSDs are created equal, be they NVMe or SATA, and some drives definitely have poorer performance than others (and particularly under certain conditions), but for general use I don't think you'd see any real-world benefit to using a 970 EVO over a good SATA SSD. Your money, your choice, of course, but this is definitely an area where you could shave a chunk off the cost without compromising on actual noticeable performance.

    Then again, if you've not upgraded your PC for a long time then I can definitely see the attraction of splashing out a bit on all the shiny stuff... That's what I did when I retired my previous PC two years back (ATX s775 E8500 with 4GB of DDR2 and a 256GB 830 SSD plus a 1TB rust spinner, in a Fractal R2, replaced by an mITX Z170 board and i7 6700K, 16GB of DDR4, the two SSDs mentioned above, all in a lovely NCase M1). Could I have saved a lot of cash by dialling down a couple of those items? Sure, but where's the fun in that? :D
     
  9. Gareth Halfacree

    Gareth Halfacree WIIGII! Lover of bit-tech Administrator Super Moderator Moderator

    Joined:
    4 Dec 2007
    Posts:
    17,132
    Likes Received:
    6,728
    The key real-world scenario I've seen is loading the game Exapunks, which I was playing on both the laptop and desktop. On the desktop, it took a good minute or so to load; on the laptop, about ten seconds. Now, there are other differences between the two - the laptop has a Core i7-7560U, the desktop the aforementioned A10-5800K - so I can't be sure that it's all down to the difference in transfer rate, but it certainly won't be hurtin'.
    Yeah, what I don't want to do is cut a corner now and then not have the money to correct it in the future - this needs to last me a good five or six years, just like the old one did.

    Added bonus: if I do get a new drive - whether it's SATA or NVMe - I can buy a £30 case-and-PSU bundle and have my old system ready to run as a backup if something dies on the new build.
     
  10. bawjaws

    bawjaws Multimodder

    Joined:
    5 Dec 2010
    Posts:
    4,284
    Likes Received:
    891
    Aye, if I had to guess then I'd say the CPU is playing a greater part in the difference than the storage drive. Not to say that the latter isn't making any difference, just that it's not going to be the major factor. Games don't tend to benefit too much from really fast SSDs as typically you're loading a lot of relatively small files so you'll struggle to hit full throughput - but then again, this is exactly the sort of situation where SSDs thrive relative to rust spinners :D

    As an example from my experience, loading a game like GTAV or Skyrim from my NVMe drive isn't noticeably faster than from my SATA SSD. The difference is maybe a single-figure percentage, whereas on paper the NVMe is capable of more than five times the read speed. On the other hand, loading either of these games from the 850 EVO is massively, massively quicker on my current PC than it ever was from the exact same drive on my C2D machine!

    It's funny, though, because when you're speccing yourself a new PC it's incredibly easy to justify spending x because of feature a. Once you've done that, it's easy to justify spending x+y because of feature b, and so on :D I also think that us PC nerds love to focus on features rather than benefits: it's great having all of these on-paper features but spending money to have something that's good on paper but doesn't actually benefit you in reality is just throwing money away, really.
     
  11. Vault-Tec

    Vault-Tec Green Plastic Watering Can

    Joined:
    30 Aug 2015
    Posts:
    14,973
    Likes Received:
    3,735
    Once I'm into Windows I can't tell the difference between any of my ssds for which there are many of all brands and types (OCZ, SanDisk, Samsung and Intel)

    Hence happy medium. It's faster SATA but £60 cheaper.
     
  12. Gareth Halfacree

    Gareth Halfacree WIIGII! Lover of bit-tech Administrator Super Moderator Moderator

    Joined:
    4 Dec 2007
    Posts:
    17,132
    Likes Received:
    6,728
    My desktop:
    Code:
    blacklaw@trioptimum:~$ sysbench --num-threads=4 --test=cpu run
    sysbench 0.4.12:  multi-threaded system evaluation benchmark
    
    Running the test with following options:
    Number of threads: 4
    
    Doing CPU performance benchmark
    
    Threads started!
    Done.
    
    Maximum prime number checked in CPU test: 10000
    
    
    Test execution summary:
        total time:                          1.5819s
        total number of events:              10000
        total time taken by event execution: 6.3233
        per-request statistics:
             min:                                  0.60ms
             avg:                                  0.63ms
             max:                                 15.53ms
             approx.  95 percentile:               0.70ms
    
    Threads fairness:
        events (avg/stddev):           2500.0000/24.77
        execution time (avg/stddev):   1.5808/0.00
    A teeny-tiny ickle Nvidia Jetson AGX Xavier:

    Code:
    nvidia@jetson-0423718017663:~$ sysbench --num-threads=4 --test=cpu run
    sysbench 0.4.12:  multi-threaded system evaluation benchmark
    
    Running the test with following options:
    Number of threads: 4
    
    Doing CPU performance benchmark
    
    Threads started!
    Done.
    
    Maximum prime number checked in CPU test: 10000
    
    
    Test execution summary:
        total time:                          1.4269s
        total number of events:              10000
        total time taken by event execution: 5.6954
        per-request statistics:
             min:                                  0.56ms
             avg:                                  0.57ms
             max:                                  1.57ms
             approx.  95 percentile:               0.58ms
    
    Threads fairness:
        events (avg/stddev):           2500.0000/1.58
        execution time (avg/stddev):   1.4238/0.00
    BUT WAIT, THERE'S MORE: The Jetson actually has eight cores, not four...

    Code:
    nvidia@jetson-0423718017663:~$ sysbench --num-threads=8 --test=cpu run
    sysbench 0.4.12:  multi-threaded system evaluation benchmark
    
    Running the test with following options:
    Number of threads: 8
    
    Doing CPU performance benchmark
    
    Threads started!
    Done.
    
    Maximum prime number checked in CPU test: 10000
    
    
    Test execution summary:
        total time:                          0.7585s
        total number of events:              10000
        total time taken by event execution: 6.0432
        per-request statistics:
             min:                                  0.55ms
             avg:                                  0.60ms
             max:                                  7.13ms
             approx.  95 percentile:               0.64ms
    
    Threads fairness:
        events (avg/stddev):           1250.0000/5.98
        execution time (avg/stddev):   0.7554/0.00
    Yeah, I'm definitely overdue an upgrade... (Though, in fairness, the Jetson is, like, three times the price of my PC, so there's that to take into account.
     
  13. Gareth Halfacree

    Gareth Halfacree WIIGII! Lover of bit-tech Administrator Super Moderator Moderator

    Joined:
    4 Dec 2007
    Posts:
    17,132
    Likes Received:
    6,728
    Well, the money I would need to do this has done arriven - the only question is, pull the trigger or wait for Zen 2?

    Advantages of waiting:
    Zen 2 will probably be better.
    Zen 1.5 will definitely be cheaper, at least for a bit.
    I don't have to find time to actually build the bugger.

    Advantages of doing it now:
    Zen 1.5 has had all the Linux-related bugs ironed out, which won't be the case for Zen 2 at launch.
    I won't fritter the money away on something else.
    It'll help bring down my next payment on account to HMRC.

    Decisions, decisions...
     
  14. sandys

    sandys Multimodder

    Joined:
    26 Mar 2006
    Posts:
    4,932
    Likes Received:
    727
    Do it now, you have an A10 absolute power has not been a concern for you, Ryzen is more than twice the single core performance of a high clocked A10 and of course the multicore is leaps ahead as it has proper multithreading and more cores which your use model can take advantage of. Makes sense to go with a known quantity rather than the new stuff if you want it to just work.

    FYI I submitted some A10 numbers on a Cinebench R20 thread, because I'm fun like that :D Here so you can see how it looks up against modern stuff.

    My A10 is a 5xxx series but to my knowledge AMD did not move the CPU cores on much for the APU.
     
  15. Gareth Halfacree

    Gareth Halfacree WIIGII! Lover of bit-tech Administrator Super Moderator Moderator

    Joined:
    4 Dec 2007
    Posts:
    17,132
    Likes Received:
    6,728
    How important is having Samsung B-die DRAM? I ask, 'cos finding 'em is proving an absolute pain. There's this handy tool, but the bulk of what it recommends ain't in stock in the UK - and what is in stock is around twice the price of the 3,000MHz C16 stuff I threw into my basket earlier in the thread, which is... y'know, not great.
     
  16. sandys

    sandys Multimodder

    Joined:
    26 Mar 2006
    Posts:
    4,932
    Likes Received:
    727
    People are running all sorts of RAM, 2xxx is less picky than 1xxx as a lot of AGESA/BIOS work has gone on to iron out Ryzen issues, but it is always prudent to see if you can find those sticks on a QVL of any board you are looking at, B Die is good for high clocks and low latency but Ryzen will run most memory.
     
  17. Gareth Halfacree

    Gareth Halfacree WIIGII! Lover of bit-tech Administrator Super Moderator Moderator

    Joined:
    4 Dec 2007
    Posts:
    17,132
    Likes Received:
    6,728
    Good to know, ta - I can live with losing a few percent absolute performance and saving ~£150, given that I'm going to be gaining a bajillion percent compared to my current system!
     
  18. adidan

    adidan Guesswork is still work

    Joined:
    25 Mar 2009
    Posts:
    19,800
    Likes Received:
    5,591
    Boards have been known to be picky (Ryzen 1 moreso than 2 as @sandys says) so checking your RAM is on the compatible list is top of the 'things to do'.

    Is it worth the price premium? It wouldn't be for me. So long as it's 3000 or 3200 mhz.
     
  19. veato

    veato I should be working

    Joined:
    15 Jan 2010
    Posts:
    874
    Likes Received:
    193
    I was concerned about RAM and Ryzen with talk of B-die, dual/single rank and whatnot but in the end I found a good (read cheap) deal on some 3200 RAM and decided to just go with it. It's running perfectly stable at 2933 (I can't be arsed to spend time tweaking and pushing it to its fastest overclock) and it's not like I'm sat around thinking my PC is slow.

    FWIW it's G.Skill Ripjaws, no idea about the die, *think* it's dual rank (which my mobo technically only officially supoprts upto 2400), possibly not on the QVL list and it's paired with a Ryzen 2600 and B450 ITX board.
     
  20. IanW

    IanW Grumpy Old Git

    Joined:
    2 Aug 2003
    Posts:
    9,207
    Likes Received:
    2,716
    From what I've read elsewhere, BIOS updates for B450/X470 boards have started landing in preparation for Ryzen 2.
    (Also, I'm seeing reports that the Threadripper 1900X is down below £280 now)
     

Share This Page