Processors (CPU)

Published on May 8th, 2017 | by GaK_45


Ryzen Infinity Fabric

Investigating the Performance Impact of Ryzen Infinity Fabric

With the release of AMD’s ‘Zen’ based Ryzen series of CPUs a very interesting side-effect has occurred in the PC marketplace. Namely what was once simply marketing spin has become a potential reality. You see a dirty little ‘secret’ of the PC industry is that RAM speed has very little impact on overall performance compared to putting that extra money towards a better CPU, GPU, SSD. So much so that for most consumers opting for anything much above what the CPU manufacturer designed the CPU around resulted in minor improvements. Oh sure synthetic benchmarks told a different story, but the net improvement for opting for uber-high performance, and uber-high priced, RAM is minor to say the least.

AMD’s Ryzen CPU series actually changes all that… and makes RAM speed possibly critically important. This is because AMD made a very interesting design decision and tied their Infinity Fabric speed to the RAM speed. Specifically the Infinity Fabric runs at 1:2 clock speeds of the RAM (or to be even more specific at 1:1 of the IMC’s speed setting as DDR4 is double data rate and ‘DDR4-2600’ is actually running at 1300mhz not 2600mhz as the name suggests). Since the Infinity Fabric is how the two CCX (4 core CPU Complex building block) talk to each other and the rest of the system the higher the transmission speed, the higher the bandwidth… and theoretically the higher the overall performance. This is because the shorter the period of time one block of CPU cores are waiting for data the happier they will be.

Today we are going to investigate exactly how big a deal this really is and what speed of RAM you should opt for… and put some of the savings over an Intel HEDT system towards. To do this we are going to use three different sets of 32GB (4x8GB) DDR4 RAM. Option 1 is going to HyperX Fury DDR4-2133 RAM (model HX421C14FB2K4/32), option 2 will be HyperX Fury DDR4-2400 (HX424C15FBK4/32), and the third is going to be HyperX Fury DDR4-2666 (HX426C15FBK4/32). We have chosen theses three sets of RAM as they are good value kits we have used in the past, offer great capacity, and look sexy as all get out.

Equally important all three sets of DDR4 RAM are 1.2v 4x8GB kits with good timings of C15 or better – so they are fast for their class but not so fast as to require much in the TLC department. Put another way these are low hassle, ‘plug and play’ kits that should just plain work while not skimping on capacity. As for prices, 32GB of HyperX Fury 2133 will set you back about $220 USD, the 2400 kit will cost you $250 USD, and the 2666 kit is still a very reasonable $290. While this is ‘only; a $70 variance across all three sets it is almost enough of a difference but a Crucial MX300 275GB drive which costs about $90 (USD).

We have stopped at DDR4-2666 as beyond this people have to choose between capacity and speed… and honestly anything above DDR4-2666 is a bit of a crapshoot when it comes to working with Ryzen processors and large capacity RAM sticks. So let’s warm up our kits, getting our Ryzen 1800X humming along… and see what the actual numbers are and if the added -albeit still very reasonable – expense is worth it or not.

About the Author


"Knowledgeable, opinionated and not afraid to ask the questions you can’t or won’t." GaK_45's combination of multiple industry certifications(MCSE, CCNA, various CompTIA, etc), and over twenty years' experience in the computer industry allows him to provide detailed analysis that is as trustworthy as it is practical.

  • David Anderson

    Running the fabric at ram clock speed ensures that they don’t miss a chance to read or write and have to wait for the next available cycle. To avoid large time penalties, they have to maintain a whole multiplier ratio between fabric and ram, i.e. 1:1, 2:1, or 3:1. Any other ratio leads to asynchronous timing and missed read/write cycles. At 2:1, the fabric internal clock would have to be 2400MHz, which may or may not be needed with the current arrangement. I sure would like to see a 16 core/32 thread version with that kind of interconnect speed. I would also like to see a more flexible FSB speed beyond 100MHz become standard too.

    • gak_45

      Yes it was indeed a design choice by AMD to make it sync and not async. Intel went with async for their cache/uncore ‘bus’ (eg 7700k is 4.5ghz cpu / ddr4-2400 / 4.2ghz uncore). IMHO a missed cycle would be well worth the OPTION to increase or decrease this critical components ratio to allow increased stability with higher clocked ram kits.

      Basically its an optics issue right now… as people wont see a missed cycle or twenty… but they do see a ram kit that wont post.

      Yes motherboard mfg’ers are working closely with AMD to make this a non-issue for most people. Which is awesome. 🙂

  • Asak

    Thanks for testing, however I have a couple of criticisms. First of all, the graphs with a non-zero based horizontal axis tends to exaggerate the differences observed. For example, on GTA V the difference is about 2 fps or 3.5%, but the graph makes it look like the difference is massive. The graphs being drawn that way doesn’t end up telling us anything because no matter how small the difference is, the charts adjust to make it seem big. Since the visual cue isn’t doing anything, you might as well skip the graph entirely and report only text based results. A better solution would be to simply start the horizontal axis at zero, instead of having it adjust to the data.

    Secondly, it would have been nice to see some faster speeds like 3200 MHz / 1600 MHz IF. Getting 2400 MHz or 2666 MHz memory seems like a no brainer, but as you get up to 3000 or 3200 MHz and tighter memory timings the price goes up a lot. So, seeing how those faster speeds compare would be interesting.

    • gak_45

      Thank you for your feedback! We always love to hear what people have to say. 🙂

      The reason for lower speed but higher capacity RAM is we fell the impact of RAM speed on real world performance is minor, whereas having enough ram is critical. Right now Ryzen has issues with 3000+ kits of ram that are bigger than 16GB. As this (or if this) changes we will revisit as needed. Basically its a crap shoot… and we would not want to inadvertently steer people towards mega-speed kits just because WE got it to work. Instead these are more – as you put it – ‘no brainer’ speeds that anyone can work with.

      The charts were not set as zero as there IS a difference but it would have hidden it almost entirely. The difference is not being exaggerated, it is being highlighted – as that difference is the point of the article. Basically it makes it easier to see what we are talking about.

      • gak_45

        No its not a true monolithic design. Simple litmus test. Can all cores talk directly to each other without the use of a cross-bar? Can all cores share cache without first transfering (copying) the data from one cluster to the other? Is there a (relatively) large performance penalty with inter-core communications from cluster to cluster?

        Thus it is not a true monolithic design. Thus our original comment and point.

        Once again this was a design choice by AMD. There are pros and cons to their choice. Overall we believe they made the right choice for most people. Not all, but no design is perfect and can please all people all the time. Understand their design and if its merits line up with your needs it is a no-brainer.

        YMMV mate.

  • Forymndare

    Can you please test with 16gb 3200? This is a possible real world setup.

    • gak_45

      Can’t make any promises but we will see what we can do 🙂

  • davidcianorris

    This article is being quoted in a missleading way because of your first sentence about “Non monolithic design”… I understand in general what you meant with that but the reality is that Ryzen is a monolithic chip and design trends had varied with time greatly. I was wondering if it wasn’t better to call it by what it really is: a modular design, instead of adding unnessesary complexity to the article. For me in short “Non traditional monolithic design” = Modular design, could that be de case if so?

    • gak_45

      That certainly is one way of putting things. The fact is AMD has gone to a design that relies upon different blocks of ‘cpus’ to scale up the core count. It is why the IF speed issue is an issue for some. In our opinion it is not a monolithic core as it has two separate and distinct ‘blocks’ that are connected by a cross bar. That is not ‘non-traditional monolthic’ design. This is only true for the big Ryzens and the small single CCX.

      Please do not get the wrong idea, its not a bad thing. Its not a slight against AMD. Its just a thing. Some will care, some will not. At the end of the day the design does not matter all that matters is how it performs and its price. By opting for a multi-CCX approach AMD can offer much better bang for the buck… and adapt the design quickly to meet new demands. It is the way of GPUs so its foolish to think that monolithic is the be all and end all for CPUs.

      • davidcianorris

        “In our opinion it is not a monolithic core as it has two separate and distinct ‘blocks’ that are connected by a cross bar.”

        But this was precisely my point… This is not a matter of opinion, regardless of the design choices (crossbars, modules, etc) the core is monolithic because its made from a single die. Monolithic means made out from a single stone thus Ryzen is monolithic. On the other hand Epyc is a proper MCM, a non monolithic design.

Back to Top ↑