Core to Core to Core: Design Trade Offs

AMD’s approach to these big processors is to take a small repeating unit, such as the 4-core complex or 8-core silicon die (which has two complexes on it), and put several on a package to get the required number of cores and threads. The upside of this is that there are a lot of replicated units, such as memory channels and PCIe lanes. The downside is how cores and memory have to talk to each other.

In a standard monolithic (single) silicon design, each core is on an internal interconnect to the memory controller and can hop out to main memory with a low latency. The speed between the cores and the memory controller is usually low, and the routing mechanism (a ring or a mesh) can determine bandwidth or latency or scalability, and the final performance is usually a trade-off.

In a multiple silicon design, where each die has access to specific memory locally but also has access to other memory via a jump, we then come across a non-uniform memory architecture, known in the business as a NUMA design. Performance can be limited by this abnormal memory delay, and software has to be ‘NUMA-aware’ in order to maximize both the latency and the bandwidth. The extra jumps between silicon and memory controllers also burn some power.

We saw this before with the first generation Threadripper: having two active silicon dies on the package meant that there was a hop if the data required was in the memory attached to the other silicon. With the second generation Threadripper, it gets a lot more complex.

On the left is the 1950X/2950X design, with two active silicon dies. Each die has direct access to 32 PCIe lanes and two memory channels each, which when combined gives 60/64 PCIe lanes and four memory channels. The cores that have direct access to the memory/PCIe connected to the die are faster than going off-die.

For the 2990WX and 2970WX, the two ‘inactive’ dies are now enabled, but do not have extra access to memory or PCIe. For these cores, there is no ‘local’ memory or connectivity: every access to main memory requires an extra hop. There is also extra die-to-die interconnects using AMD’s Infinity Fabric (IF), which consumes power.

The reason that these extra cores do not have direct access is down to the platform: the TR4 platform for the Threadripper processors is set at quad-channel memory and 60 PCIe lanes. If the other two dies had their memory and PCIe enabled, it would require new motherboards and memory arrangements.

Users might ask, well can we not change it so each silicon die has one memory channel, and one set of 16 PCIe lanes? The answer is that yes, this change could occur. However the platform is somewhat locked in how the pins and traces are managed on the socket and motherboards. The firmware is expecting two memory channels per die, and also for electrical and power reasons, the current motherboards on the market are not set up in this way. This is going to be an important point when get into the performance in the review, so keep this in mind.

It is worth noting that this new second generation of Threadripper and AMD’s server platform, EPYC, are cousins. They are both built from the same package layout and socket, but EPYC has all the memory channels (eight) and all the PCIe lanes (128) enabled:

Where Threadripper 2 falls down on having some cores without direct access to memory, EPYC has direct memory available everywhere. This has the downside of requiring more power, but it offers a more homogenous core-to-core traffic layout.

Going back to Threadripper 2, it is important to understand how the chip is going to be loaded. We confirmed this with AMD, but for the most part the scheduler will load up the cores that are directly attached to memory first, before using the other cores. What happens is that each core has a priority weighting, based on performance, thermals, and power – the ones closest to memory get a higher priority, however as those fill up, the cores nearby get demoted due to thermal inefficiencies. This means that while the CPU will likely fill up the cores close to memory first, it will not be a simple case of filling up all of those cores first – the system may get to 12-14 cores loaded before going out to the two new bits of silicon.

The AMD Threadripper 2990WX 32-Core and 2950X 16-Core Review Precision Boost 2, Precision Boost Overdrive, and StoreMI
Comments Locked


View All Comments

  • MrSpadge - Monday, August 13, 2018 - link

    I don't think AVX512 is going to matter much anytime soon. However, The 8 memory channels of EPYC could matter a lot for HPC.
  • ElFenix - Monday, August 13, 2018 - link

    You guys need a 4k or maybe even 5k workload for transcoding - it's thread limited at 1080p so it becomes IPC and turbo limited. With x265 you can load up multiple 1080p handbrake instances on these high core count processors and they don't break a sweat.
  • ElFenix - Monday, August 13, 2018 - link

    That should be *1080p spawns limited numbers of threads*
  • T1beriu - Monday, August 13, 2018 - link

    >Europe is famed for its lack of air conditioning everywhere

    UK is a lot colder generally in the summer compared to the rest of Europe. I wouldn't generalize the lack of AC for the rest of Europe. AC is pretty common in my country.
  • jospoortvliet - Saturday, August 18, 2018 - link

    Missing everywhere here in Germany... though after this insanely hot summer i bet that that will begin to change...
  • powerincarnate - Monday, August 13, 2018 - link

    I didn't see a lot of gaming benchmarks, which I guess I understand, since these are more workstation cpus. It would be good to have seen both though to get a better idea of the overall qualities of the cpu as a multipurpose care.

    It seems from tomshardware benches that 7980xe, especially when overclocked, is best overall. AMD 2990wx obviously winning on the pure multi-threaded workstation stuff as long as it is not memory intensive.

    It seems like the 2950x from both of these sites, is really the processor to get from the threadripper lineup.

    it seems when gaming is taken to account, the best of both worlds is the 7900x

    And for gaming, and when you factor price as well, the 8700k, 8086, and slightly behind, the 2700x are the cards to get.

    Overall.... I'm a little disappointed in this release. Was much more impressed with the 2700x. It's likely since we didn't really get a true change in the manufacturing process or design of the chip, that the limitations of the 2990wx will probably be ironed out with Zen2 (this is Zen+ after all).
  • bill.rookard - Monday, August 13, 2018 - link

    Looking at it myself, yeah - these really aren't gaming CPUs by any stretch of the imagination, thus the lack of gaming benchmarks is perfectly understandable to me. As for the results of the benchmark results? I'm thinking the 2950x is the sweet spot. Lower power, lower latency, more power for the cores vs interconnects, and a much higher clockspeed makes it IMHO the better choice unless you have those fringe workloads which requires a bunch-o-cores.
  • shendxx - Monday, August 13, 2018 - link

    this guy come from Toms that said 7900x is best for both world, lol, when the graph from toms show clearly even on gaming, 2950x is equal on Minimum FPS with 8700k and only lose 3 to 10 FPS On Average,
  • apoclypse - Monday, August 13, 2018 - link

    I don't know. Gaming performance is the least thing I care about with this chip but that seems be all most tech press cares about, especially tech tubers. These chips are not for gaming. If anything these chips should be compared to Intel's Xeon line as it seems that is actually where AMD is aiming these at since they don't have a dedicated SKU for workstation chips like Intel. These are only marketed as HEDT chips because it gives AMD positive press, but if anything the ones who should be paying attention should be OEM high end workstation builders. In that regard Threadripper is more than compelling. It's higher clocked than Intel's Xeon chips, has more cores for less money, and still has all the pro level features that is needed for workstation level work.

    I think AMD should lean into that a bit more in their marketing but that stuff isn't sexy and it doesn't grab attention like marketing it towards rich and stupid "gamers", and the technorati who eat that stuff up.

    This is a workstation chip period, and should be treated, tested and benchmarked as such, imo.
  • Icehawk - Monday, August 13, 2018 - link

    If only the tier 1 vendors would offer TR workstations... I really wanted to purchase a few for work to use as VM hosts but my only real option is Xeon currently. The 32 core monster would likely make for a great VM host for mid-weight usage.

Log in

Don't have an account? Sign up now