A Year Later
How does AMD’s 2nd generation of Ryzen CPUs stack up to Intel’s Coffee Lake?
Despite what might be considered an overall slump in enthusiast PC building due to record low GPU availability and sky-high memory prices, 2017 was one of the most exciting and competitive years in recent history when it comes to CPU innovation. On the desktop side alone, we saw the launch of AMD's new Zen CPU architecture with the Ryzen 1000 series of parts starting last March; we also saw new HEDT platforms from both Intel and AMD, and Intel's first 6-core mainstream CPUs.
Although the timeline doesn't quite work out for Ryzen to have affected the engineering-side of Intel's decision to release a 6-core desktop processor, it's evident AMD's pressure changed Intel's pricing and release schedule.
With little desktop competition, it's likely that the i7-8700K would have been a more expensive part, and released later. It's likely that Coffee Lake would have seen a full stack product launch in early 2018, as opposed to the staggered launch we experienced where only one compatible chipset and a subset of CPUs were available for months.
AMD and Ryzen have put significant pressure on Intel to remain competitive, which is good for the industry as a whole.
We're now at just over a year since AMD's first Ryzen processor releases, and looking at the first appearance of the codename Pinnacle Ridge CPUs. Launching today are the Ryzen 7 2700X and 2700, and the Ryzen 5 2600x and 2600 processors. Can AMD keep moving the needle forward in the CPU space? Let's take a look.
First, let's get this out of the way. Despite the Ryzen 2000-series branding, the Pinnacle Ridge processors do not mark the appearance of the Zen 2 architecture, nor any other significant architectural-level changes.
Instead, Pinnacle Ridge follows more of the tick-tock model that Intel processor generations traditionally adhered to. The Ryzen 2000 series is a slightly tweaked Zen architecure, but manufactured this time on GlobalFoundries new 12nm process instead of the previous 14nm. Combined, this archituecture tweaks and process node change are being referred to by AMD as "Zen+."
|Ryzen 7 2700X||Ryzen 7 1800X||Ryzen 5 2600X||Ryzen 5 1600X||Core i7-8700K||Core i5-8600K|
|Architecture||Zen+||Zen||Zen+||Zen||Coffee Lake||Coffee Lake|
|Base Clock||3.7 GHz||3.6 GHz||3.6 GHz||3.6 GHz||3.7 GHz||3.6 GHz|
|Boost Clock||4.3 GHz||4.0 GHz||4.2 GHz||4.0 GHz||4.7 GHz||4.3 GHz|
|TDP||105 watts||95 watts||95 watts||95 watts||95 watts||95 watts|
Editor's Note: A previous version of this specs table incorrectly listed the TDP of the Ryzen 5 2600X as 65W due to a confusion from the AMD reviewer's guide. The TDP has been updated in this chart to reflect the correct value of 95W.
Changes from the Ryzen 7 1800X to the Ryzen 7 2700X include a 100MHz bump in both base and boost clocks, as well as support for DDR4-2933 memory. The TDP of the 2700X also jumps 10W compared to last year's flagship AM4 processor.
The Ryzen 5 2600X gains a 200MHz bump on the boost clock and DDR-2933 memory support while maintaining the same 65W TDP as the R5 1600X.
The most significant change to the processor design with this generation comes in the form of latency improvements in the memory system. I'll be blunt here: AMD didn't give us much detail on HOW these changes were made, and instead only left us with a few specific claims:
AMD tells us that L1 cache latency is 13% improved, L2 cache sees the biggest boost with a 34% improvement, and the L3 cache latency improves by 16%. Along with that, main system DRAM gets an 11% improvement. All of this is stated in "up to" terms, so it appears this isn't a global change but one that depends on the workload. The result is a 3% IPC improvement.
You might remember that we spent quite a bit of time last year looking into the latency questions surrounding the Zen architecture, from Ryzen to Threadripper. We used a "ping tool" to find the thread-to-thread communication times of the Ryzen parts, facilitating a bit more understanding of what the underlying architecture was doing.
Obviously, we wanted to do the same here for the Ryzen 2700X.
The gray and the yellow lines represent the Ryzen 1800X and the Ryzen 2700X at a fixed clock speed of 3.6 GHz and running at a DDR4 memory speed of 2667 MHz. I wanted to get a clear clock-for-clock change in the latencies we measured. The green line bumps up the memory speed to 2933 MHz, the highest JEDEC memory speed for the 2700X.
Here's what is interesting: the latency on the inter-CCX threads (thread 2 to thread 7) indicates that the 2700X is SLOWER than the 1800X when running at the same memory speed and clock speed. The CCX-to-CCX latency is FASTER on the 2700X, however, going from ~130ns to ~120ns. The performance improvement on the last layer of latency does appear that it is more valuable to the overall performance of the system than the inter-CCX latency (that goes from 50ns up to 55ns).
But clearly, the performance change on the CCX-to-CCX comms didn't come without a cost in other areas of the chip. But the balance likely weighed heavier in favor of this shift than leaving the architecture as-is. Again, AMD didn't share details about this change, so we are just guessing at the REASONS for the results we are measuring.
Interestingly, when we bump the memory speed up to 2933 MHz, the new JEDEC speed supported by the 2700X, the inter-CCX latency of the 2700X matches that of the 1800X and nets us ANOTHER 8-10ns of latency reduction on the CCX-to-CCX latency.
We are doing more diving on this moving forward, but for now, that's what we're left with.
With a new processor launch, comes a new chipset. For the Pinnacle Ridge processor release, the newest AM4 chipset is the X470. The X470 chipset itself doesn't have any dramatic changes over its X370 predecessor, except for the addition of native USB 3.1 Gen 2 directly from the chipset.
However, the launch of the X470 is a great opportunity for motherboard vendors to release revised AM4 boards, taking what they've learned in the past year about the platform and building upon it. One such example of this is the ASUS Crosshair VII Hero board we used for our review testing.
The X470 chipset also marks the launch of NVMe RAID for the AM4 platform, as well as StoreMI caching technology. These features will be also be added to at least some previous AM4 boards through UEFI updates. For more information on these storage technologies, you can check out Allyn's quick look at storage performance.
This time around, all Ryzen 2000-series processors will come with a stock cooler option in the box. Previously, the higher-end products such as the Ryzen 7 1800X required users to bring their own cooling solution. In the early days of the AM4 socket, this proved a bit of a challenge finding a compatible cooler.
While the AM4-enabled cooler ecosystem is much more robust now, stock coolers can be a good option for users not looking to overclock their system.
With the Ryzen 7 2700X comes the new Wraith Prism, RGB LED-enabled cooler.
In the box with the Ryzen 5 2600X, users will receive the Wraith Spire cooler previously bundled with the last generation Ryzen 5 processors.
|Review Terms and Disclosure
All Information as of the Date of Publication
|How product was obtained:||The product is on loan from AMD for the purpose of this review.|
|What happens to the product after review:||The product remains the property of AMD but is on extended loan for future testing and product comparisons.|
|Company involvement:||AMD had no control over the content of the review and was not consulted prior to publication.|
|PC Perspective Compensation:||Neither PC Perspective nor any of its staff were paid or compensated in any way by AMD for this review.|
|Advertising Disclosure:||AMD has purchased advertising at PC Perspective during the past twelve months.|
|Affiliate links:||This article contains affiliate links to online retailers. PC Perspective may receive compensation for purchases through those links.|
|Consulting Disclosure:||AMD is a current client of Shrout Research for products or services related to this review.|