A large number of column inches have been put towards describing and explaining AMD's new underlying scalable interconnect: the Infinity Fabric. A superset of HyperTransport, this interconnect is designed to enable both the CPUs and GPUs from AMD to communicate quickly, at high bandwidth, low latency, and with low power with the ability to scale out to large systems. One of the results of the implementation of Infinity Fabric on the processor side is that it runs at the frequency of the DRAM in the system, with a secondary potential uplift in performance when using faster memory. The debate between enthusiasts, consumers and the general populous in regards to Ryzen's memory performance and has been an ever-raging topic since the AGESA 1.0.0.6 BIOS updates were introduced several weeks ago. We dedicated some time to test the effect of high-performance memory on Ryzen using Team Group's latest Night Hawk RGB memory.

Memory Scaling on Ryzen 7: AMD's Infinity Fabric

Typically overlooked by many when outlining components for a new system, memory can a key role in system operation. For the last ten years, memory performance for consumers has been generally inconseqential on memory speed: we tested this for DDR3 for Haswell and DDR4 for Haswell-E, and two major conclusions came out of that testing:

  • As long as a user buys something above the bargain basement specification, performance is better than the worst,
  • Performance tapers to a point with memory, very quickly hitting large price increases for little gain,
  • The only major performance gain that scales comes from integrated gaming

So it is perhaps not surprising to read in forums that the general pervasive commentary is that “memory speed over DDR4-2400 does not matter and is a con by manufacturers”. This has the potential to change with AMD's Infinity Fabric, where the interconnect speed between sets of cores is directly linked with the memory speed. For any workload that transfers data between cores or out to main memory, the speed of the Infinity Fabric can potentially directly influence the performance. Despite the fact that pure speed isn’t always the ‘be all and end all’ of establishing performance gains, it has the potential to provide some gains with this new interconnect design.

The Infinity Fabric (hereafter shortened to IF) consists of two fabric planes: the Scalable Control Fabric (SCF) and the Scalable Data Fabric (SDF). 

The SCF is all about control: power management, remote management and security and IO. Essentially when data has to flow to different elements of the processor other than main memory, the SCF is in control.

The SDF is where main memory access comes into play. There's still management here - being able to organize buffers and queues in order of priority assists with latency, and the organization also relies on a speedy implementaiton. The slide below is aimed more towards the IF implementation in AMD's server products, such as power control on individual memory channels, but still relevant to accelerating consumer workflow.

AMD's goal with IF was to develop an interconnect that could scale beyond CPUs, groups of CPUs, and GPUs. In the EPYC server product line, IF connects not only cores within the same piece of silicon, but silicon within the same processor and also processor to processor. Two important factors come into the design here: power (usually measured in energy per bit transferred) and bandwidth.

The bandwidth of the IF is designed to match the bandwidth of each channel of main memory, creating a solution that should potentially be unified without resorting to large buffers or delays.

Discussing IF in the server context is a bit beyond the scope of what we are testing in this article, but the point we're trying to get across is that IF was built with a wide scope of products in mind. On the consumer platform, while IF isn't necessarily used to such a large degree as in server, the potential for the speed of IF to affect performance is just as high.

AGESA 1.0.0.6 (aka AGESA 1006) and Memory Support

At the time of the launch of Ryzen, a number of industry sources privately disclosed to us that the platform side of the product line was rushed. There was little time to do full DRAM compatibility lists, even with standard memory kits in the marketplace, and this lead to a few issues for early adopters to try and get matching kits that worked well without some tweaking. Within a few weeks this was ironed out when the memory vendors and motherboard vendors had time to test and adjust their firmware.

Overriding this was a lower than expected level of DRAM frequency support. During the launch, AMD had promized that Ryzen would be compatible with high speed memory, however reviewers and customers were having issues with higher speed memory kits (3200 MT/s and above) . These issues have been addressed via a wave of motherboard BIOS updates built upon an updated version of the AGESA (AMD Generic Encapsulated Software Architecture), specifically up to version 1.0.0.6.

Given that the Ryzen platform itself has matured over the last couple of months, now is the time for a quick test on the scalability on AMDs Zen architecture to see if performance can scale consistency with raw memory frequency, or if any performance gains are achieved at all. For this testing we are using Team Group's latest Night Hawk RGB memory kit at several different memory straps under our shorter CPU and CPU gaming benchmark suites.

Recommended Reading

Team Group's Night Hawk RGB Memory: 2x8GB of DDR4-3000 CL16
Comments Locked

65 Comments

View All Comments

  • HStewart - Wednesday, September 27, 2017 - link

    I think of this way, AMD was desperate to get back into business of CPU's, but financially they had some issues to really think it though. So they created an 8 Core Zen and then clunk them together so that they claim higher core count. This designed is likely primary why it does not scaled.

    But they did something that they probably didn't want to here - they ignored that Intel has been making higher core CPU's in the Xeon and that it quite simple for them to place them in gaming machines. This has a good side effect for Intel owners, because it means it keeps Intel on its toes - but the bad news I am afraid is that AMD will not be financially able to keep up with core wars and eventually have to drop - also purchasing ATI has alienated potential buyers - who in right mind would purchase an AMD GPU on Intel CPU.

    One thing that is interesting, is that Intel and the industry is moving in a different direction. Mobile is where the industry is going not huge fat desktops. This is a place where AMD is missing the mark and could possible complete loose it enter company open and solely based there efforts on the desktop industry.
  • duploxxx - Thursday, September 28, 2017 - link

    lol dude what have you been smoking?

    A) its intel that responded on AMD core count
    B) Zen 8 core multi die was in the design from the start to keep cost low
    C) Xeon v2 and v3 both had issues with scale out on core hence the reason for the new grid which is sub optimal on caching.
    D) Intel has a way more expensive die, you forget that they ask 2500+ euro upto 14000 euro for there 16+ cores? while AMD charges 4000 euro for 32 cores. The gold series dont even come close to AMD offerings in cores.
    E) Intel is not moving at all, they own the biggest part of the industry on x86 and that is what they try to keep. THey lost the low power war vs ARM and they sure try to get into IOT with lots of money but it aint that easy.
    F)AMD has low budget so they infiltrate markets where they believe they can gain.
  • cap87 - Friday, September 29, 2017 - link

    Intel didn’t respond to AMD with higher core counts, processors are designed years in advance, suggesting otherwise is just plain ignorance. What they did do was push forward the release date of Coffee Lake thanks to AMD’s pressure.
  • jospoortvliet - Saturday, September 30, 2017 - link

    Their design was meant for servers. Bringing the 18 cores suddenly to the high end desktop was most certain my something they kept as an ace option but it wasn't their original plan and that is obvious from the way it was rushed to market, being months later than the 10 core model and many of the earlier motherboards barely or simply not able to handle the load. They are also obviously clocked very high with barely room for overclock and breaking their tdp, throttling under heavy use on many boards even without oc.
  • Hixbot - Monday, October 9, 2017 - link

    I'm tired of hearing this. what you are suggesting is ignorance.
    Intel had loads of time to R&D, yes designs take years, but they've had those years to design coffe lake with 4 cores, with 8 cores. They design and design, they could have designed a sandy bridge as 8 core and not released it. You think they need to take years to respond to a competitive push. Let me tell you, they can design all sorts of options "years in advance" and only bring to market what they choose. So if it weren't for Zen, we might be staring at a 4 core (max) coffee lake. OMG it's hilarious to see this "design takes years" argument. They can and do take years to design all sorts of potential processors, they can then choose what to bring to market in a much shorter time.
  • Arbie - Wednesday, September 27, 2017 - link

    For what? Some small and very expensive ST performance increase? Consider what AMD has done for us in reigniting competition and moving the tech envelope forward. Think what that took, and whether they could possibly do it again. Anyone who doesn't absolutely have to buy Intel this time around should give the nod to AMD. They've earned it, where Intel has not. Really, the tech is almost equal and in most regards AMD gives you more for the dollar. If we as consumers don't respond to that, vigorously, they may give up. How would you like an Intel-only future?
  • Nagorak - Thursday, September 28, 2017 - link

    Plenty of other tests have shown significant scaling. This is with loose subtimings. You can get even more performance from tight subtimings on top of faster memory speed. Remember Ryzen was only about 8% slower clock for clock than Kaby Lake. Faster memory speeds make up most of that difference, albeit Ryzen can't run at such high frequency as KL.
  • notashill - Wednesday, September 27, 2017 - link

    I'm curious if the higher clocked parts scale any better, presumably they were spending more time waiting on memory in the first place. The tests were done with a 1700, 1800X has 20% higher all-core clock.
  • willis936 - Wednesday, September 27, 2017 - link

    It would seem that 2 channels of DDR4 is not enough to keep 8 cores fed. It will be interesting to see if it's enough to keep 6 cores fed on coffee lake since intel's memory subsystem is higher performance but they also have higher single threaded performance (and may need more memory throughput as a result).
  • sor - Wednesday, September 27, 2017 - link

    “AGESA 1.0.0.6 BIOS updates were introduced several weeks ago”

    Shouldn’t that be several *months* ago, or was there some more recent AGESA release from the one being discussed in April/May?

Log in

Don't have an account? Sign up now