It’s here, it’s finally here. The months of hype, guessing, heated debates; AMD finally opens up the floodgates for benchmarks to be revealed on its latest CPU launch, Ryzen. The boost in instructions per clock, the doubling of core counts, smaller manufacture process, new architecture, new socket, new I/O, it’s all new, and we can finally reveal how it actually performs!
Well… we would if there wasn’t a huge delay in the delivery of our review samples. It pains us to say it, but we don’t have any launch day benchmarks to share with you today. Our box of Ryzen CPUs, literally arrived yesterday, and Rob is only just getting back from GDC. Frustrating would be an understatement, but it is what it is. However, that’s not to say we won’t be doing a review, no sir. We just have to work feverishly over the next week to get something up for you all.
With that said, we can go into some of the background developments of the Ryzen CPU that haven’t been covered so far, as well as a quick look at upcoming Naples server-based CPUs.
The table above details the three CPUs coming out right now, all of which are 8-cores 16 threads. To clarify, these are indeed real cores, unlike the old Bulldozer and Excavator cores that were split integer with a shared floating point module.
In the original press releases, the initial performance gains of Zen over the previous gen were put at roughly 40%. AMD has updated this to a 52% improvement in IPC (Instructions Per Clock), meaning getting more done for the same clock speed – per core.
This was made possible by a number of changes in the architecture, which includes better branch prediction (requesting data from memory before it’s needed), better control over clocks (more steppings), smaller and faster transistors, changes to the chip layout, as well as tighter integration of the independent subsystems (such as memory controller and IO).
It’s the price points that are going to be the real decider here. It’s been said before with the 1800X going toe-to-toe with Intel’s Core i7-6900K, for half the price. Even if this is under specific workloads that are highly thread dependent, that’s still impressive. Unfortunately, many tasks are still dependent on single thread performance, with one critical sector of importance being games, and for many people, this will still be the deciding factor.
However, if there was one thing we learnt while at GDC, it’s that AMD is already way ahead on this problem – getting games out of this single-thread rut that’s been plaguing us for over a decade. While there has been a lot of focus on Zen and Vega, AMD is also trying to tie the two together in software and through developer access to low-level APIs such as DX12 and Vulkan.
The launch of Ryzen is more than just a CPU refresh for AMD, it’s the beginning of a shift; moving away from dual and quad-core mainstream processors to 8-core. Once these chips start circulating, it’s going to force developers to change focus.
AMD did provide some in-house test results for its three new CPUs under various gaming conditions. You can see those in the slides above. However, once we get our rig setup, we’ll be going through and verifying them ourselves. Even at first glance, Intel is still beating AMD under a number of tests, so the price-point equivalents are still viable for certain workloads.
Ryzen though, is more about core counts, and it’s those tests that really shine for AMD; it’s a shame those mostly boil down to media encoding and rendering… at least for now. This does indirectly have a major benefit for gaming though, and that’s streaming. While all the big players are adding video encoding engines into their respective GPUs, the options for those encoders are still limited. CPU encoders can get much better quality than GPU, and are less dependent on drivers, so for the most part CPU encoding is still prefered for streaming. Ryzen with all those extra cores to spare means people can game and stream with pretty much no impact on game performance.
Ryzen had another surprise in store too, the fact that all of the chips will support overclocking. No black editions here, every single Ryzen CPU can be overclocked with a compatible motherboard. This is where the second surprise crops up, it’s not just prosumer boards that can overclock, even the sub-$100 boards can overclock these CPUs with either a B350 or X370 Chipset. Even the X300 SFF boards can overclock.
We covered some of the new features that are coming with the chipset release with the ASUS AM4 board announcements
, and this extends out to other manufacturers too, including MSI, GIGABYTE, ASRock and Biostar. The short of it is that native support for USB 3.1 Gen.2, NVMe, and PCIe 3.0 will all come as standard across the range of boards, with additional PCIe 2.0 and SATA ports provided by the chipset.
The other changes brought about by Zen extend beyond the desktop market though, and if we’re being honest, Naples is more exciting than Ryzen in some regards. AMD has been out of the server market for a long time now, but Zen is offering a new opportunity to claw back some of that market share from Intel.
The high core counts from the desktop chips are scaling up even further with Naples, 32-core CPUs that can scale almost perfectly with the 64-core dual socket boards that will be coming out too. Shared buses, a fabric of scalable interfaces between CPUs and GPUs, dedicated links between CPUs, all the sorts of things needed for virtualization and complex arithmetic.
Once we get a chance, we’ll put Ryzen through its paces over the coming week and do a proper launch article then.
For the time being, you can check out reviews by good friends of ours over at Gamers Nexus, Legit Reviews and LAN OC.