Review: AMD Ryzen 5 1600 processor - just fantastic value

Discussion in 'Frontpage news' started by Hilbert Hagedoorn, May 16, 2017.

  1. Fox2232

    Fox2232 Ancient Guru

    Messages:
    11,809
    Likes Received:
    3,366
    GPU:
    6900XT+AW@240Hz
    New i5, like yours is better if overclocked. but not for long. Older i5s are already choking on games...
    It is only question of time till all 4cored CPUs take same road. And that time is pretty close. Luckily they will still handle 60fps well. And therefore will remain good for most of users.
    But if one talks about 60fps, then i5 is not better than any Ryzen, because they are same or worse as Ryzen deals better with additional workloads like recording+encoding or preloading of data.
     
    Last edited: May 17, 2017
  2. Amx85

    Amx85 Master Guru

    Messages:
    333
    Likes Received:
    10
    GPU:
    MSI R7-260X2GD5/OC
    X370 Krait + R5 1600 sounds like my next setup lol

    nice review
     
  3. G4ry4Caesar

    G4ry4Caesar Banned

    Messages:
    10
    Likes Received:
    0
    GPU:
    IGP
    Ryzen is problematic. It has a huge ~110ns ping delay between threads communicating across the CCX interconnect fabric and that's if you're using fast 3200MHz RAM. A thread accessing shared cache that's not in its own CCX takes 2 to 3 times as long.

    Modern games spawn several simultaneous often interdependent threads that share data and migrate across CPU cores all the time. Having massive delays when one thread is crossing to the next CCX or trying to access data that resides in the shared cache of the second CCX is extremely inefficient in this type of workloads.

    Why did AMD release 2-CCX chips? Games don't need 6 or 8-core CPUs. 100% of games are made to run on consoles which are powered by AMD's low-power low-performance Jaguar CPUs, and then ported to PCs. In fact, the best gaming CPU at the moment is not Intel's 10-core CPU, it's Intel's 4-core Kaby Lake CPU.

    Had they released a single CCX properly clocked 4-core Ryzen, they would've been doing much better now.
     
  4. PrMinisterGR

    PrMinisterGR Ancient Guru

    Messages:
    7,895
    Likes Received:
    767
    GPU:
    Inno3D RTX 3090
    It's just a design choice, and actually not detrimental at all. Most of the performance differences we see with Kaby are due to higher frequencies and compiler settings. You also forget to mention that although Intel parts do indeed have the same latency between cores, that latency is 2x the latency that Ryzen cores have within a CCX.

    [​IMG]

    [​IMG]

    Ryzen CPUs within the CCX have a max 40ns latency, while Intel cores are at a constant +50% latency at 80ns. Only when CCX to CCX communication happens Ryzen parts get 140ns latency.
     

  5. G4ry4Caesar

    G4ry4Caesar Banned

    Messages:
    10
    Likes Received:
    0
    GPU:
    IGP
    Actually cross-CCX latency is what causing Ryzen to underperform in many applications besides games. Also, latency between Intel Kaby Lake CPU cores is roughly 40ns too. Check out the Tomshardware review.
     
  6. PrMinisterGR

    PrMinisterGR Ancient Guru

    Messages:
    7,895
    Likes Received:
    767
    GPU:
    Inno3D RTX 3090
    I just posted the PCPer results that neither AMD nor Intel have disputed. You have no idea what's causing what to underperform, and neither do I. We both suppose. But with news like simple LLVM setting changes to "Haswel" instead of Bulldozer giving Ryzen 5-10% performance boosts, I'm more inclined to believe that the main issue at this point is codepaths and not the CCX.
     
  7. sykozis

    sykozis Ancient Guru

    Messages:
    22,008
    Likes Received:
    1,197
    GPU:
    MSI RX5700
    If it wasn't for the memory compatibility issue, I'd move from my 6600K to an R5 1600. That said... I was curious as to why there are no results from the 6600K included in the review charts.

    My 6600K won't clock beyond 4.3ghz.... Not sure if it's the motherboard or processor.
     
  8. Aura89

    Aura89 Ancient Guru

    Messages:
    8,189
    Likes Received:
    1,286
    GPU:
    -
    Why are so many random people creating accounts lately only to post in AMD related forums to bring up topics that have already been discussed multiple of times and generally either figured out to not be an issue, be incorrect or possibly an issue but can't be pinned down, but has been discussed to death already and has no reason to keep being brought up? (Unless something ACTUALLY NEW happens)

    What's the point of your post G4ry4Caesar? What was stated in it that hadn't been discussed to death already? What benefit could it possibly bring to the conversation other then to flamebait? And why did you seemingly create an account just to do that?

    And yes i see your one and only other post in this forum.

    3 posts, all to say "ryzen sucks, and here's why i don't know what i'm talking about"

    Why? What's the point?
     
    Last edited: May 18, 2017
  9. RandomDriverDev

    RandomDriverDev Banned

    Messages:
    111
    Likes Received:
    0
    GPU:
    1080 / 8GB and 1060 / 6GB
    hmm... FMA3 hardlocks, VMx extention seg faults.

    these chips weren't ready for release, they should have held off a few more months.

    This isn't exactly a fact.
    some Games scale to all physical cores on 6 core processors, but suffer on 4 core chips with hyperthreads.

    NVidia's DX11 driver scales up to and runs best on 6 core processors, which is part of the reason it shows a throughput loss on the 1500 and 1400 processors - NVidia's mid-low level optimisations don't understand that AMD's design suffers when shared resources cross the CCX.

    I do agree that this CCX design should have been kept to server grade releases, and consumers grade should have opted for 4c/8th.

    That said, the CCX interconnect does appear it can be scaled up in the future for higher throughput at lower clocks.
     
    Last edited: May 18, 2017
  10. PrMinisterGR

    PrMinisterGR Ancient Guru

    Messages:
    7,895
    Likes Received:
    767
    GPU:
    Inno3D RTX 3090
    FMA3 hardlocks

    It's also an AGESA issue, not a CPU issue, and it's getting bios patched. Even Skylake (the 6th generation of an established architecture) had crashes with Prime95, and there are still PCIe issues with X99.
    :infinity:

    Haven't seen any issues about VM extensions, do you have any links?
     
    Last edited: May 18, 2017

  11. RandomDriverDev

    RandomDriverDev Banned

    Messages:
    111
    Likes Received:
    0
    GPU:
    1080 / 8GB and 1060 / 6GB
    It is a CPU issue - any hardlock is a result of a register getting stuck either through invalid data or physical flaw with the execution pipe

    the microcode will work around it with a possible performance hit.

    http://www.os2museum.com/wp/vme-broken-on-amd-ryzen/

    basically cannot run Windows 2k or XP with hardware extensions in use.
     
    Last edited: May 18, 2017
  12. Aura89

    Aura89 Ancient Guru

    Messages:
    8,189
    Likes Received:
    1,286
    GPU:
    -
    By that definition, no major new processor is ever ready to be released, since all, including intel, have problems with major new architectures at first.

    So i guess we'll just never receive a new architecture, sounds good right?
     
  13. PrMinisterGR

    PrMinisterGR Ancient Guru

    Messages:
    7,895
    Likes Received:
    767
    GPU:
    Inno3D RTX 3090
    I didn't know or imagined that a problem like that would exist. It's interesting to see it's more or less fixable with disabling the capability from the VM itself, but it's a bit of an edge case no? Also if you read the comments on the article itself, most people believe this (like the FMA3 issue) are basically microcode issues and not hard locks.
     
  14. kapu

    kapu Ancient Guru

    Messages:
    5,008
    Likes Received:
    529
    GPU:
    Radeon 6800
    Was talking about 7600K , all of them clock 4.8Ghz minimum. So it's only fair to compare overlocked i57600k vs Ryzem which doesnt oveerlock at all.

    I found the benchs

    [​IMG]
    [​IMG]

    Its exmaple how HUGE difference this simple OC does ... and YOU buy K version for OC ....
     
    Last edited: May 18, 2017
  15. vbetts

    vbetts Don Vincenzo Staff Member

    Messages:
    15,128
    Likes Received:
    1,703
    GPU:
    GTX 1080 Ti

    I think people are getting too obsessed with overclocking...

    If the core is efficient and works well on its own, and cannot overclock as high as the competition but still remains efficient with the small overclock it can achieve, then how is that a valid point? Sure, you can get 4.8ghz to 5ghz on the 7600 from 3.8 ghz, but despite that change in the clocked frequency you aren't gaining a whole lot over the 1600X being a whole 1ghz lower in clock.

    Aside from the fact those results are from Tom's Hardware which has been known for years to favor Intel and will do what they can to show that, I don't see why you don't use Guru3d graphs anyway. You are on Guru3d forums after all.
     

  16. PrMinisterGR

    PrMinisterGR Ancient Guru

    Messages:
    7,895
    Likes Received:
    767
    GPU:
    Inno3D RTX 3090
    Do the benches show CPU usage during gameplay? That matters too.
     
  17. kapu

    kapu Ancient Guru

    Messages:
    5,008
    Likes Received:
    529
    GPU:
    Radeon 6800
    There are no benchmarks like this on guru. Where do you see OCed i5 vs Ryzen ?

    It is more than a valid point when both cpus cost the same. You get 20-30% performace free on 7600k just by changing multi, even child could do that :).
     
  18. PrMinisterGR

    PrMinisterGR Ancient Guru

    Messages:
    7,895
    Likes Received:
    767
    GPU:
    Inno3D RTX 3090
    You have to adjust cache frequencies and find optimal voltages also. There is also the simple fact that no matter how fast those four threads are (and they aren't that much faster individually), you're purchasing a four thread CPU versus a twelve thread one. A Ryzen 5 + RX 580/570 are most likely the best choices in this segment, why would you ever get a quad again?
     
  19. vbetts

    vbetts Don Vincenzo Staff Member

    Messages:
    15,128
    Likes Received:
    1,703
    GPU:
    GTX 1080 Ti
    It would help to know if these results show cpu usage, that's why it's not a valid point still. IPC performance, Intel might be better at that but then when you consider the 1600 is $30 less and has 2 more cores and 4 more threads compared to the 7600k and overclocks as well as an X chip does without needing anything special, it's a no brainer.

    The approach AMD is taking with Ryzen is the way it should be, clocked frequencies are not as high but efficiency is what they are looking for and it shows. What's the point of going from 3.8ghz to 5ghz when the scaling on it isn't the greatest? Don't get me wrong it's not bad, but at this point it's just saying I've got the bigger number here.
     
    Last edited: May 18, 2017
  20. kapu

    kapu Ancient Guru

    Messages:
    5,008
    Likes Received:
    529
    GPU:
    Radeon 6800
    Not you dont, i haven't changed a single setting just the multi.

    Point wasn't quad vs six or octa core , but i5 vs Ryzen 5 for gaming. Both are great and quite smilar , and i pointed it is not black and white which is better, which is why i wanted highlight OC potential in i5 7600k - Can't argue that there is 20-30% free performance you can tap in to with simple tweak.
     

Share This Page